BOUDIN, Circuit Judge.

London Egemonye was indicted in 1993 under a multi-count indictment charging him and others with conspiracy and other offenses relating to the possession and use of other people's credit cards. 18 U.S.C. §§ 1029(a)(2) (trafficking, fraud and use), 1029(a)(3) (possession with intent to defraud), 1029 (b)(2) (conspiracy). On June 10, 1994, Egemonye entered guilty pleas to all counts, and he now appeals from his sentence arguing that it is flawed by the government's manipulation of sentencing factors and by an improper computation of loss.

Because there was no trial, we derive the facts primarily from the recitations at the plea hearing, from the presentence report, and from submissions at the sentencing hearing. United States v. Connell, 960 F.2d 191, 192-93 (1st Cir. 1992). The case arose out of a sting operation conducted by a joint federal-state task force investigating credit card and other financial fraud in Massachusetts. The critical events took place in January and February 1993.

Robert Leslie, who was cooperating with authorities, introduced Egemonye to an undercover state trooper known to both only as "Kathy." On January 21, 1993, Kathy supplied Egemonye with two BayBank MasterCard credit cards and one BayBank Visa credit card with an aggregate credit limit of $7,450 for all three cards. Egemonye then created false driver's licenses in the credit card names, each license bearing Leslie's photograph, and drove Leslie to three different banks to obtain cash advances of $6,900.

Egemonye purchased four more credit cards from Kathy on January 29, 1993, and four more on February 2, 1993. The aggregate limits on the cards in the two transactions were $21,000 and $14,000, respectively. In between these transactions, several of the cards were used to obtain advances from banks, and Egemonye and others in the conspiracy engineered deposits of some stolen checks into accounts of individual card holders to boost the depleted credit available for those cards.

Until the fourth transaction, Kathy made the "sales" in exchange for a share of the proceeds, but on February 5 she proposed that she be paid a flat $200 per card. Egemonye said, "I'm not going to buy one card for two hundred. . . . It has to be like ten." On February 10, Kathy told Egemonye that she expected to receive a number of cards that day, that Egemonye should bring $2,000 for 10 cards, and that she would "front" (finance) any additional cards and accept payment for them later. Egemonye agreed, subject to his examination of the cards.

When Kathy and Egemonye met later that day, Kathy said that she had a bag full of cards and asked Egemonye whether he knew of another buyer if he did not want them all. He said, "I probably can handle them," and proceeded to give Kathy $2,000 down, and a promise of $6,000 more later, for 40 Household Bank Visa and MasterCard credit cards with an aggregate limit of $200,000. Egemonye was arrested immediately thereafter, followed by the indictment and plea already described.

At sentencing, the district court increased the base offense level of 6 by 8 additional levels because the "loss" attributed by the court to Egemonye was over $200,000. U.S.S.G. §§ 2F1.1(a), (b)(1)(H). The court computed the loss at $242,950, representing the aggregate credit limit of the 51 credit cards purchased from Kathy in the four transactions. The offense level was then adjusted in other respects, not here in dispute, and Egemonye was sentenced within the guideline range to 37 months' imprisonment.

1. On appeal, Egemonye's first claim is directed at the 40 cards supplied to him in the final transaction. Egemonye contends that including these 40 cards in the loss calculation condones "blatant sentencing factor manipulation engaged in by the investigating agents" and is a violation of constitutional due process. He relies on several decisions, including United States v. Connell, 960 F.2d 191, 196 (1st Cir. 1992).

We have recently had occasion to discuss Connell and the other decisions in this circuit that have addressed sentencing factor manipulation. United States v. Montoya, 62 F.3d 1 (1st Cir. 1995). Summarizing the prior cases, we said that "where government agents have improperly enlarged the scope or scale of the crime," the sentencing court has power to exclude "the tainted transaction" from the guideline computations and for purposes of any mandatory minimum statute. Montoya, 62 F.3d at 3 (quoting in part Connell, 960 F.2d at 195).

However, recognizing the broad latitude allowed to the government in investigating and suppressing crime, we stressed that it was only "extraordinary misconduct" by agents that could give rise to such an exclusion, which would occur in the teeth of a statute or guideline approved by Congress. Montoya, 62 F.3d at 3-4 (quoting in part United States v. Gibbens, 25 F.3d 28, 31 (1st Cir. 1994)). While something less than a constitutional violation might suffice, as extraordinary misconduct, Egemonye's reference to due process concepts is certainly in the ballpark.

In Montoya, as in previous cases, we refused to lay down fixed rules to define sentence factor manipulation, but said that the focus is normally upon the conduct of the government rather than the defendant. 62 F.3d at 4. Indeed, Egemonye does not claim that his will was overborne or deny that he was predisposed to the offense. What Egemonye claims is that the fourth transaction had no legitimate law enforcement purpose and was designed solely to boost his federal sentence because government agents were unhappy with lenient treatment that Egemonye earlier received in state court.

There is some basis for the suggestion that task force agents were unhappy with Egemonye's prior record and believed, in the words of one of the agents, "that he [earlier] got off lightly for his criminal activity." That criminal record, according to the agent just quoted, involved a history of credit card fraud by Egemonye that could be traced back to 1990 and involved a number of transactions. On this appeal, the government is prepared to assume arguendo that the background facts, "viewed collectively, could call the government's motives into question to some extent."

Nonetheless, the government says that multiple sales were clearly appropriate in order to identify Egemonye's co-conspirators, which they did. As to the final sale of 40 cards, the government insists that it too "had a valid investigatory purpose" which was "to explore the parameters of the defendant's criminality." Egemonye's counsel replies that this "parameters" explanation has no real substance and could be used to enlarge a defendant's sentence to virtually any height whatever. We think that Egemonye's reply has some force but overstates the matter.

There is, it should be stressed, no indication that Egemonye was coerced or pressured to achieve a new level of crime. True, the fourth sale was much larger than the earlier ones; but agent Kathy did not force the 40 cards on Egemonye. On the contrary, he had insisted on at least 10 cards for the new $200 per card payment ("I'm not going to buy one card for two hundred. . . . It has to be like ten.") And when offered a bag full of cards — with the request that he recommend another buyer for those he did not want — he responded, "I probably can handle them," and took them all.

Government agents are not limited to replicating a suspect's largest unsolicited crime. In this case, the full contours of the criminal operation — its size, techniques, personnel — were, like an iceberg, largely submerged; and the means of exploration were additional and larger transactions. The first three transactions clearly served this purpose and the fourth, even though followed immediately by the arrest, provided air-tight evidence for trial that Egemonye was a significant dealer and not a petty swindler. While the sting could not be endlessly prolonged and enlarged, nothing in the objective facts suggests "misconduct" at all, let alone "extraordinary misconduct."

The question, then, is whether the fourth transaction is tainted by the agents' subjective motives. The pallet in such matters contains not blacks and whites but shades of gray. Motives may be mixed; good and bad motives are often matters of degree; and there can be multiple actors. Whether to consider subjective motive at all presents a problem of policy. Compare Harlow v. Fitzgerald, 457 U.S. 800, 102 S.Ct. 2727, 73 L.Ed.2d 396 (1982) (refusing to do so in the qualified immunity context). Still, we would be greatly concerned if evidence otherwise available showed that a plainly improper subjective motive — say, racial hostility or personal animus — had enlarged or prolonged the sting.

But this is not such a case. About the most that can be derived from the record, drawing all reasonable inferences in favor of Egemonye, is that the agents thought that Egemonye was an established and unrepentant defrauder who had escaped serious punishment for a series of past, similar frauds. With this in mind, they conducted a sting operation that involved no pressure whatever on Egemonye, lasted for only four transactions, and garnered several other defendants. The first three transactions involved 11 cards; the last one, 40. This is a sizeable jump but hardly extraordinary.

That agents considered Egemonye's past record in selecting him for overtures by the task force is a commonplace of law enforcement. Undercover operations frequently target those who are suspected of crime, and the recent history of fraudulent activities gave the agents some reason to think that Egemonye was not only predisposed but actively engaged. Fed.R.Evid. 404(a), restricting character evidence to show propensity, is a rule for trials and not the conduct of police investigations.

At worst, the agents went too far if and to the extent that they thought themselves entitled to make up for any shortfall in prior punishments. But the line is thin and blurred between such a dubious motive and a simple desire to be sure that a committed criminal is caught and tried for a substantial offense based on unshakeable evidence. And, as we have already held, Egemonye was legitimately targeted and the sting objectively reasonable in extent. Under these circumstances, even assuming that the agents' motives were mixed and not of crystalline purity, we see nothing that would require a curtailment of the sentence.

2. Egemonye's second challenge to his sentence concerns the district court's computation of loss. As already noted, the governing guideline keys the offense level primarily to "the loss" caused by the offense, U.S.S.G. § 2F1.1(b)(1) (loss table), but goes on to provide ( id., comment (n. 7)) that intended loss should be used if it is greater than actual loss:

Consistent with the provisions of § 2X1.1 (Attempt, Solicitation or Conspiracy), if an intended loss that the defendant was attempting to inflict can be determined, this figure will be used if it is greater than the actual loss . . . For example, if the fraud consisted of selling or attempting to sell $40,000 in worthless securities. . . . the loss would be $40,000.

In accord with the presentence report, the district court in this case attributed to Egemonye an intended loss equal to the aggregate limits of the purchased credit cards. A reading of the transcript indicates that the judge found that Egemonye was capable of and intended to use the cards to secure amounts at or virtually at their aggregate limits. We review such a factual determination only for clear error, United States v. Pavao, 948 F.2d 74, 77 (1st Cir. 1991), reserving for closer scrutiny a buried legal issue shortly to be described.

On the factual issue of intended use and capability, the government bears the burden of proof because an increase in the offense level was sought, see United States v. Sklar, 920 F.2d 107, 112 (1st Cir. 1990), but the guideline itself cautions that a reasonable estimate of loss will suffice. U.S.S.G. § 2F1.1 comment. (n. 8). Egemonye begins by pointing out that he realized only about 53 percent of the aggregate card limits from the cards involved in the first three transactions and nothing at all from the final bagful of cards since he was apprehended almost immediately. He argues that to predict a 100 percent recovery is simply unrealistic.

Unfortunately for Egemonye, there was affirmative evidence that he instructed his runners at the outset to procure cash from the banks at or virtually at the card limits. In addition, he arranged for the deposit of stolen checks into some of the accounts, in order to refresh their limits. By this means, some of the accounts could have been milked for amounts in excess of their aggregate limits. The 53 percent figure represented only the amount that Egemonye had secured at the time his scheme was interrupted by arrest. See United States v. Strozier, 981 F.2d 281, 284 (7th Cir. 1992).

In sum, taking the issue purely as a factual one of intent and capability, we do not think that on this record the use of the aggregate card limits as a measure of intended and potential loss was clearly erroneous. Where there is good evidence of actual intent and some prospect of success, we do not think that a court needs to engage in more refined forecasts of just how successful the scheme was likely to be. See United States v. Lorenzo, 995 F.2d 1448, 1460 (9th Cir.), cert. denied, ___ U.S. ___, 114 S.Ct. 225, 126 L.Ed.2d 180 (1993). The situation may be quite difficult where intent must be inferred solely from the likely effects of the scheme. See United States v. Stern, 13 F.3d 489 (1st Cir. 1994).

But there is a wrinkle. There is a cross-reference in U.S.S.G. § 2F1.1's application note 7 (quoted above in pertinent part) to U.S.S.G. § 2X1.1; and there is a second such cross-reference in application note 9, which reads (in pertinent part):

"In the case of a partially completed offense (e.g., an offense involving a completed fraud that is part of a larger, attempted fraud), the offense level is to be determined in accordance with the provisions of § 2X1.1 . . . whether the conviction is for the substantive offense, the inchoate offense . . ., or both."

Egemonye's counsel argues that section 2X1.1, and the discount it makes available, apply in this case.

U.S.S.G. § 2X1.1 is concerned with determining the offense level for an attempt or conspiracy; and this it sets at three levels less than the offense level for the substantive offense — unless the defendant (or his co-conspirators) have completed all of the acts believed necessary for the substantive offense or were "about to complete all such acts" when apprehended. For cases within the "unless" clause — which the background comment says represent "most" cases — there is no such discount. Effectively, the guideline gives the defendant a three-level discount if he is some distance from completing the substantive crime.

Read literally, section 2X1.1 is not relevant to the present case because 14 of the 15 counts against Egemonye involved completed substantive offenses, ranging from trafficking in unauthorized credit cards to producing false driver's licenses, and the conspiracy thus embraced fully completed crimes. On the other hand, the cross-reference to section 2X1.1 in section 2F1.1 arguably connects the intended loss concept to the attempt guideline, and section 2X1.1 blurs the matter further with the following application note (comment. (n. 4)), providing (in pertinent part):

In certain cases, the participants may have completed . . . . all of the acts necessary for the successful completion of part, but not all, of the intended offense. In such cases, the offense level for the count (or group of closely related multiple counts) is whichever of the following is greater: the offense level for the intended offense minus 3 levels . . . or the offense level for the part of the offense for which the necessary acts were completed. . . . For example, where the intended offense was the theft of $800,000 but the participants completed . . . only the acts necessary to steal $30,000, the offense level is the offense level for the theft of $800,000 minus 3 levels, or the offense level for the theft of $30,000, whichever is greater.

Interpreting these provisions is a matter of some difficulty, and the only cases in point are in conflict. Compare United States v. Watkins, 994 F.2d 1192 (6th Cir. 1993) with United States v. Strozier, 981 F.2d 281 (7th Cir. 1992). The problem, in a nutshell, is that section 2X1.1 has on its face nothing to do with a completed substantive offense or a conspiracy that has been carried to completion. On the other hand, the notion of a discount could be extended from the case of an incomplete offense to that of a completed offense where intended harm is part of the calculus and the harm is only partly completed.

Recognizing the question to be close, we are inclined to stand by the literal language of the guidelines that directs section 2X1.1 to cases where the substantive offense has not been completed. E.g., United States v. Sung, 51 F.3d 92 (7th Cir. 1995). The argument for a discount for inchoate crimes is obvious; the defendant has started down the road toward the substantive crime but has not gotten there yet and, whatever his intention, might still turn back before crossing the line. By contrast, Egemonye did cross the line and commit the substantive crime by acquiring the cards and making the false documents, so the basic purpose of the section 2X1.1 discount has nothing to do with him.

Where a completed offense is involved, it is surely rational to measure culpability in part by the intended harm and to refuse a discount where the offense is complete even though the intended harm has not yet been fully realized. From the standpoint of moral guilt, and dangerousness, there is little to distinguish such a defendant from one who has actually inflicted the same amount of harm. And we are influenced in part by the fact that the case law calculating sentences based on intended harm, most of it admittedly without making reference to section 2X1.1, is consistent with this view. E.g., United States v. Guyon, 27 F.3d 723 (1st Cir. 1994) United States v. Resurreccion, 978 F.2d 759 (1st Cir. 1992).

Of course, there would be nothing irrational in deciding that actual harm is worse than intended harm and providing a three-level discount wherever the sentence for a completed offense is measured in part by intended harm. But this is not in general the philosophy of the guidelines; if it were, possession of drugs with intent to distribute would be punished less harshly than the actual sale of an equivalent amount. The wrinkle of section 2X1.1 cannot be ironed completely smooth, but the pertinent language already quoted can in fact be squared with our result.

Thus, the cross-references in section 2F1.1 are easily explained; they do invoke the discount, or the possibility of a discount, where the underlying crime is merely an attempt or conspiracy. Application note 4 in section 2X1.1 is less easily reconciled; but we think the difference is that in the theft case, there is no completed crime as to the larger amount but only (in substance) an attempt. Here, by contrast, all 51 of the cards were the subject of completed crimes.

Egemonye's remaining claim as to loss is to argue that no consideration should be given to the 40 cards in the fourth transaction, or at least to the unexpected 30 cards (over and above the ten cards Egemonye requested). This is largely a restatement of the claim that sentencing factor manipulation occurred. Having rejected that claim, we think that — from the standpoint of intended loss — Egemonye can fairly be charged with intending to inflict loss as to all of the cards.

Both issues in this case are difficult and important. We are thus especially indebted to counsel for the able briefing and argument presented on both sides. The Sentencing Commission's attention will be drawn to the arguable lack of clarity in the interplay between section 2F1.1 and section 2X1.1.

Affirmed.