Opinion for the Court filed by Circuit Judge WILLIAMS.
STEPHEN F. WILLIAMS, Circuit Judge:
After a jury trial, Chevalier Thompson was convicted on one count of distribution of cocaine base and on one count of possession with intent to distribute cocaine base. We find no merit in his one claim of trial error— that the government used an improper “identification procedure” by showing witnesses a photograph of Thompson in the course of its immediate pre-trial preparation. Although we affirm the convictions, we remand for further consideration of his sentence.
I. Alleged Use of an Improper Identification Procedure
On January 3, 1992, undercover police officers Leslie Russell and Donita Giles approached a man, later identified as Thompson, and indicated their interest in buying $20 worth of cocaine base. Thompson took the undercover officers to another person, Anthony Carter, and told Carter to “hold them”. Thompson then got into the driver’s
seat of a burgundy Nissan automobile and reached down for something. When he came back, he produced a black film cannister that contained some Ziploe bags. He handed a bag to Officer Russell in exchange for a prerecorded $20 bill, and told Russell that if she wanted any more, she should “just come back and ask for Nissan.”
To preserve their ability to make undercover buys in the future, the officers did not themselves arrest the two men. Instead, Officer Russell broadcast a lookout for the pair. Minutes later police stopped a group of three men, two of whom matched the descriptions in the lookout. The undercover officers then drove by slowly and Officer Russell identified Carter and Thompson as the men who had been in on the sale. A search revealed that Thompson was carrying the pre-recorded $20 bill and a set of Nissan keys; the keys fit a burgundy Nissan Maxi-ma that was parked a few vehicles away from the scene of the transaction. Twenty-eight Ziplocs containing a total of 7.225 grams of cocaine base were found in the fuse box, located at knee level to the left of the steering column. The police also found, on the sidewalk where the men had been stopped, a black film cannister with 2.71 grams of cocaine base distributed among 13 Ziplocs.
At trial, both Russell and Giles gave in-court identifications of Thompson as the man who had sold Russell' the cocaine base. Thompson now argues, however, that a new trial is necessary because these in-eourt identifications should never have been admitted. On cross-examination, both officers acknowledged that they had seen Thompson’s arrest photograph in preparation for their testimony; though Russell was not asked for the surrounding details, Giles testified that the prosecutor had shown her the picture and that she had not seen any other photographs on that occasion. Thompson contends that this constituted an unduly suggestive identification procedure, creating such a risk of misidentification at trial that due process requires the reversal of his convictions. See generally
Neil v. Biggers,
409 U.S. 188, 196-201, 93 S.Ct. 375, 380-383, 34 L.Ed.2d 401 (1972).
Though the relevant facts had been elicited in cross-examination of the government’s witnesses on the first day of trial, Thompson did not raise this argument in the trial court until after the jury found him guilty, when he filed a motion for a new trial. Thompson nonetheless argues that we should review the issue as
if
he had raised a timely objection, because the trial court ruled on the issue in the context of the post-verdict motion. For purposes of determining our standard of review of an alleged error in admission of evidence, however, a post-verdict motion for a new trial is not the same as a timely objection: the delay eliminates any chance that the judge could correct the error without a duplicative trial, and according review as if a timely objection had been raised virtually invites strategic behavior by defense counsel. See, e.g.,
United States v. Saro,
24 F.3d 283, 287 (D.C.Cir.1994). Thus we review only for plain error. See, e.g.,
United States v. Breque,
964 F.2d 381, 387 n. 7 (5th Cir.1992). In fact, the scope of review makes no difference here, as we do not think that the admission of the in-eourt identifications was error at all, let alone the “obvious” and “prejudicial” error required under plain-error standards. See generally
United States v. Olano,
— U.S. -, -, 113 S.Ct. 1770, 1778, 123 L.Ed.2d 508 (1993).
It is quite a stretch to portray what happened during pre-trial preparation as an identification procedure. Russell, in fact, had made a drive-by identification of Thompson minutes after he had been arrested, well before the allegedly suggestive photo display. Under fairly similar circumstances, we have found “nothing improper” in later showing such a witness a photograph to refresh her recollection. See
United States v. Marshall,
511 F.2d 1308, 1311 (D.C.Cir.1975). To be sure, this technique may well make the witness seem more confident when she identifies the defendant in court. But the same can be said of
every
technique used to refresh a witness’s recollection during pre-trial preparations.
In any event, Russell’s sighting of the photo while preparing for her testimony carried no significant risk of causing any misidentification, and so even if it was improper
it would not preclude her in-court identification of Thompson. Cf.
Simmons v. United States,
390 U.S. 377, 384, 88 S.Ct. 967, 971, 19 L.Ed.2d 1247 (1968). In one sense, indeed, the photo gave her no information that she would not have acquired as soon as she took the witness stand at trial, for she testified on cross-examination that she knew where criminal defendants sit. More important, the key identification was the drive-by; unless Thompson could shake the validity of that, he could not raise a serious identity issue (unless the Thompson in court was different from the Thompson who was arrested, which he does not claim). And Russell’s exposure to Thompson’s arrest photo in trial preparation could have had only the most remote effect on the credibility of her testimony about the events leading up to the drive-by.
Officer Giles was not in quite the same position as Officer Russell. While Giles testified that she too had recognized Thompson during the drive-by, she said she did not
voice
her identification at the time because “[i]t was- not [her] purchase”. But this difference is immaterial unless her exposure to the photograph would in some way have been more likely to affect the credibility of her account of the drive-by than Russell’s exposure would have affected hers. We find no error in the admission of the in-court identifications.
II. Alleged Sentencing Errors
A.
Ineffective Assistance of Counsel
Thompson’s other challenges all go to the vah'dity of his sentence. Two factors worked strongly against Thompson: his offenses involved crack cocaine, which triggers much harsher punishments than powder cocaine, and he had two prior felony drug convictions. As a result of the combination of these factors, the statutory maximum for his offense was life imprisonment. 21 ' U.S.C. § 841(b)(1)(B). This statutory maximum, in turn, meant that Thompson’s offense level under the “career offender” provisions of the federal sentencing guidelines was 37. See U.S.S.G. § 4B1.1. The same provisions assign all career offenders to “criminal history” category VI, the guidelines’ highest.
Id.
The guideline range corresponding to offense level 37 and criminal history category VI is 360 months to life.
Id.
Ch. 5 Pt. A.
At his pre-sentence hearing, however, Thompson asserted that he did not learn that career-offender papers had been filed in his case until after the verdict, and so he had gone to trial under the mistaken impression that he faced a maximum sentence of 137 months under the guidelines.. (In the absence of the guidelines’ career-offender provisions, Thompson’s base offense level would have been 26,
id.
§ 2D1.1(c)(9), and his criminal history category would have been V, corresponding to a guideline range of 110 to 137 months in prison.
Id.
Ch. 5 Pt. A.) The district court ultimately accepted Thompson’s assertion, finding that “[Thompson’s] trial counsel did fail to inform the defendant that enhancement papers had been filed or that he could be sentenced to life imprisonment under the Sentencing Guidelines”. The court concluded, moreover, that this failure had violated Thompson’s Sixth Amendment right to the effective assistance of counsel: under the two-step test of
Strickland v. Washington,
466 U.S. 668, 104 S.Ct. 2062, 80 L.Ed.2d 674 (1984), the lawyer’s performance had been “deficient” and the deficiency had “prejudiced” Thompson by rendering him “unable to rationally decide whether ... to plead guilty” or to stand trial.
The court found, however, that Thompson had established only
limited
prejudice. The government had not offered him a plea bargain, and Thompson had made no proffer that he could have given the government “substantial assistance” in other prosecutions. Cf. U.S.S.G. § 5K1.1. Still, Thompson could at least have pleaded guilty to the indictment as it stood, “thereby lowering his Total Offense Level by two points” to reflect his acceptance of responsibility. See
id.
§ 3E1-.1. To reflect this possibility, the district court decided to sentence Thompson as if his offense level were 35 instead of 37 — a step that the court said would eliminate “any possible prejudice” established by Thompson. On appeal, Thompson argues that the district court undervalued his prejudice and hence devised a constitutionally inadequate remedy.
Our analysis of this issue is substantially affected by the government’s failure to challenge key parts of the district court’s logic. The government’s position, in fact, is hard to decipher. The government has not appealed the sentence, and does not contest the district court’s finding that Thompson received ineffective' assistance of counsel. But it vigorously argues that Thompson has failed to show
any
prejudice. Under the Supreme Court’s Sixth Amendment jurisprudence, these positions are internally' contradictory, at least in a technical sense. To make out a Sixth Amendment violation, the defendant must show not only that the performance of his counsel was “deficient” but also that it caused prejudice; there must be a “reasonable probability” — defined as “a probability sufficient to undermine confidence in the outcome” — that “but for counsel’s unprofessional errors, the result of the proceeding would have been different”.
Strickland,
466 U.S. at 694, 104 S.Ct. at 2068; see also
United States v. Cronic,
466 U.S. 648, 658, 104 S.Ct. 2039, 2046-2047, 80 L.Ed.2d 657 (1984).
Reconstructing the government’s argument in this light, we take it to be conceding only that Thompson has satisfied the first prong of the
Strickland
test, not that Thompson has established any prejudice within the meaning of the second prong. Still, the government says nothing at all to indicate that the district court applied the wrong legal standard in gauging prejudice. We take it for granted, then, that the district court implicitly found a “reasonable probability” that if Thompson had been properly advised, he would have pleaded guilty (thereby qualifying at least for a 2-point reduction under § 3E1.1). The government also does not argue that this factual finding was clearly erroneous. As a result of the government’s concessions, we have no occasion to review the district court’s conclusion that Thompson’s Sixth Amendment rights were violated, or its implicit conclusion that his subsequent fair trial was not an adequate remedy. We express no view on those findings or on the somewhat parallel analysis of the Third Circuit in
United States v. Day,
969 F.2d 39 (3d Cir.1992).
We are left, then, only with Thompson’s claims to greater relief. These build essentially on the proposition that there was a reasonable probability that Thompson, if he had been properly advised, would either have secured a plea bargain substantially reducing his exposure, or would at least have pleaded to the original indictment in time to secure the 3-point reduction afforded by U.S.S.G. § 3E1.1. We consider them in that order.
1.
Thompson’s Demand for a New Trial or for Sentencing at Base Level 26.
Thompson claims that if he had been properly advised about the effect of the career-offender provisions, he would have initiated plea-bargaining negotiations with the government and might well have been able to do better than simply plead “straight up” to the indictment, i.e., to both counts as charged. According to Thompson, the district court should either have granted him a new trial (to restore his bargaining power for plea negotiations) or sentenced him as if the career-offender provisions did not apply to him (to reflect the situation that he had believed to exist when he decided to stand trial).
If there is a “reasonable probability” that a properly advised Thompson would have pleaded “straight up” to the indictment, we think there must also be a “reasonable probability” that he would first have initiated negotiations to determine whether he could strike a more favorable deal. After all, if willing to plead guilty merely to get a 2-point reduction in his offense level, Thompson would certainly have been willing to plead guilty to get a
bigger
reduction. And the government has advanced no persuasive reasons why Thompson would not have investigated that possibility before settling for a mere 2-point adjustment.
But it takes two to trade; the fact that Thompson might well have initiated plea negotiations does not mean that those negotiations would have been successful. Though the district court could have been clearer about its factual findings and the legal standard that it was applying, we understand its opinion to conclude that Thompson failed to show a “reasonable probability” that he could
have struck a plea bargain under which his base offense level would have been less than 37. Despite Thompson’s arguments, we find no error in this conclusion.
Citing a new policy of the United States Attorney for the District of Columbia in cases involving less than 50 grams of crack, Thompson argues that if he were to be retried now, the trial might well occur in local court rather than in federal court; this move would reheve him from the impact of the federal guidelines and substantially reduce his sentencing exposure. But evidence that Thompson might have received more favorable treatment as a result of changes over time is weak evidence about what he might have obtained in the relevant period. It is still further weakened by Thompson’s concession that even under current policies his status as a career offender might have led to his trial in federal court.
Thompson next advances a statistic indicating that in one recent year, roughly 70 percent of all federal drug defendants pleaded guilty. This is all but irrelevant. It tells us nothing about the
terms
on which they were able to enter their guilty pleas, let alone about the specific plea-bargaining practices of the U.S. Attorney’s office for the District of Columbia at the relevant time. And even if we could assume that most defendants who plead guilty are getting more for their trouble than a 2-point reduction for acceptance of responsibility, Thompson’s statistic tells us nothing about the plea bargains that
career offenders
are able to strike.
Although Thompson describes a variety of hypothetical plea arrangements that might have been appropriate in his case, and would have greatly reduced his exposure, he offers nothing to indicate that the government would have accepted any such plea arrangement. He is thus reduced to an unsubstantiated assertion that the government surely would have agreed to
some
deal short of what the district court contemplated. Yet Thompson bears the burden of establishing the extent of his “prejudice”, and this assertion does not suffice. We obviously cannot require him to specify the exact deal that he would have been able to work out during the plea negotiations that never began. But he offers
no
information — not even the barest sort of anecdotal evidence — about how the U.S. Attorney’s office for the District of Columbia approached plea negotiations with career offenders during the relevant period. Indeed, the only information on this point comes from the government, which represented at oral argument that the office’s policy during that period would have precluded any plea arrangement that gave Thompson a base offense level of less than 37. We do not rely on this representation because it came up at oral argument rather than before the trial court (or even in the briefs), when Thompson would have had a better opportunity to respond. But the burden is Thompson’s, not the government’s, and we do not think that he has carried it.
We also reject Thompson’s argument that he was constructively denied representation altogether at a critical stage of the proceedings, with the result that prejudice should be presumed. Cf.
Strickland,
466 U.S. at 692, 104 S.Ct. at 2067. This exception is reserved for “a very narrow spectrum of eases”.
Martin v. McCotter,
796 F.2d 813, 820 (5th Cir.1986) (quoting
Chadwick v. Green,
740 F.2d 897, 901 (11th Cir.1984)); see also
Hollenback v. United States,
987 F.2d 1272, 1275 (7th Cir.1993) (exception applies only to “particularly egregious conduct that is the functional equivalent of actual absence of counsel”). For instance, our sister circuits have applied the exception when defense counsel “refused to participate in any aspect of the trial”,
Martin v. Rose,
744 F.2d 1245, 1250 (6th Cir.1984), or conceded during closing argument that the government had proved the only factual issues in dispute beyond a reasonable doubt,
United States v. Swanson,
943 F.2d 1070 (9th Cir.1991). Cf.
Javor v. United States,
724 F.2d 831, 833 (9th Cir.1984)
(pre-Strickland
case presuming prejudice where counsel slept through substantial portion of trial). Thompson’s case is by no means so egregious.
2.
Thompson’s Demand for Sentencing at Offense Level 3L
As a fallback position, Thompson argues that the district court should have reduced his offense level by one extra point to reflect
the possibility that he might have entered an
early
guilty plea if he had been properly advised. After authorizing a 2-point reduction for “the acceptance of responsibility”, the guidelines provide that defendants at as high an offense level as Thompson can reduce their offense levels by a third point if they give early notification of their intention to plead guilty, so that the government can avoid preparing for trial and the court can allocate its resources efficiently. See U.S.S.G. § 3E1.1. Noting the possibility that he could have qualified for this 3-point reduction even if the government had refused all possible plea bargains, Thompson claims that the district court should at least have given him the benefit of this provision.
The question for the district court was whether there was a “reasonable probability” that Thompson, if properly advised, would have pleaded guilty at such a time as to qualify for the 3-point reduction. The court ultimately accepted the government’s position that there was no such probability. This conclusion, on first inspection, seems to be in tension with the court’s essentially uneontest-ed finding of a reasonable probability that Thompson
would
have pleaded guilty for the sake of a mere
2
-point reduction. To the extent that a contested factual finding conflicts with an uneontested one, we would think it proper to discard the contested finding; this tendency would lead us to remand the case with instructions to award Thompson the extra point.
Here, however, the conflict between the two findings may be illusory. As the government points out, for instance, Thompson’s case featured a hotly disputed motion to suppress most of the evidence against him. See Brief for Appellee at 26 n. 15. Thompson might well have been unwilling to abandon this motion (which could conceivably have led to total dismissal of the charges against him) merely for the sake of reducing his total offense level from 35 to 34. Yet it is quite possible that Thompson would have had no chance of receiving a 3-point reduction under § 3E1.1 unless he abandoned the motion. A defendant who makes the government devote considerable resources to litigating a motion to suppress certainly has a weaker claim to the extra point than one who pleads guilty without any fight at all; in any event, trial was only two weeks away when the district court finally ruled on the motion.
Still, the district court did not make any factual findings along these fines, nor did it appear to rely on any such argument in refusing to award the third point. Instead, in making its critical finding about whether a properly advised Thompson would likely have qualified for the third point, the court apparently relied entirely on the grounds that Thompson pursued an alibi defense and continued to assert his innocence even after being found guilty. These grounds have no bearing on how information about his status as a career criminal would have affected the
timing
of his hypothetical guilty plea. At the most, they suggest that Thompson would not have pleaded guilty
at all
—a finding that the district court necessarily rejected when it concluded that the ineffective assistance of his counsel had caused Thompson some prejudice. And in any event, the choices that Thompson
actually made
do not necessarily shed any useful fight on the choices that he
would have made
if he had been properly advised.
Ordinarily, we review the district court’s factual findings only for clear error. But when it seems plain that the court arrived at those findings on the basis of faulty logic, such deference is inappropriate. Even if the ultimate findings themselves are not clearly erroneous, “the appropriate disposition of the case is to vacate the district court’s judgment and remand for further factfinding”.
United States v. Wragge,
893 F.2d 1296, 1299 (11th Cir.1990); see also, e.g.,
United States v. Harris,
959 F.2d 246, 264-65 (D.C.Cir.1992). Accordingly, we re
mand for the district court to reassess the probability that Thompson, if properly advised, would have secured the third point. In conducting this reassessment, of course, the court is free to consider all relevant facts and circumstances, including the fact that the third point would have been available to Thompson only if his sentencing had occurred after October 31, 1992.
B.
Disparate Punishment of Crack Offenses
Thompson complains that the federal sentencing guidelines punish offenses involving crack cocaine much more severely than they punish offenses involving cocaine powder. Following Congress’s lead, see 21 U.S.C. § 841(b), the guidelines treat each gram of crack as they treat 100 grams of powder, see U.S.S.G. § 2D1.1. Thompson argues that this disparity has a disproportionate impact on African-Americans, because “empirical evidence establishes that African-Americans are far more likely to be prosecuted for cocaine base offenses than are whites, and whites are far more likely to be prosecuted for cocaine powder offenses”. Appellant’s Opening Brief at 45. As the claim rests solely on the differential impact of provisions that are race-neutral on their face, and Thompson makes no suggestion that the disparity reflects any discriminatory intent, cf.
Washington v. Davis,
426 U.S. 229, 96 S.Ct. 2040, 48 L,Ed.2d 597 (1976),
the distinction is valid so long as it rests on a rational basis. Thompson claims that that basis is absent, so that the distinction violates both the due-process and equal-protection components of the Fifth Amendment. He also argues that the harsh sentences for crack offenses constitute cruel and unusual punishment under the Eighth Amendment.
As the appellant recognizes, however, we have already held that the disparate treatment of crack and powder cocaine easily survives both rational-basis review and Eighth Amendment challenge. See
United States v. Cyrus,
890 F.2d 1245, 1248 (D.C.Cir.1989). In rejecting Thompson’s position, we are in the company of all other circuits that have addressed either his Fifth Amendment claims, see, e.g.,
United States v. Williams,
982 F.2d 1209 (8th Cir.1992);
United States v. Jones,
979 F.2d 317 (3d Cir.1992);
United States v. Robinson,
978 F.2d 1554 (10th Cir.1992);
United States v. King,
972 F.2d 1259 (11th Cir.1992);
United States v. Harding,
971 F.2d 410 (9th Cir.1992);
United States v. Williams,
962 F.2d 1218 (6th Cir.1992);
United States v. Lawrence,
951 F.2d 751 (7th Cir.1991);
United States v. Galloway,
951 F.2d 64 (5th Cir.1992);
United States v. Thomas,
900 F.2d 37 (4th Cir.1990), or his Eighth Amendment claim, see, e.g.,
United States v. Frazier,
981 F.2d 92 (3d Cir.1992);
United States v. Avant,
907 F.2d 623 (6th Cir.1990);
United States v. Buckner,
894 F.2d 975 (8th Cir.1990).
Thompson also contends that even if the sentencing scheme is constitutional, the “disparate racial impact” of the guideline ranges for crack offenses provides grounds for a downward departure from those ranges. Departures, however, are warranted only if the sentencing court finds a “mitigating circumstance” that was “not adequately taken into consideration by the Sentencing Commission in formulating the guidelines” and that “should result in a sentence different from that described.” 18 U.S.C. § 3553(b). We do not think that the disparate racial impact identified by Thompson qualifies as such a circumstance.
The purpose of departures is to allow adjustment where special characteristics of the crime or the offender are so unusual as to render unjust an otherwise just sentence under the guidelines. See, e.g.,
United States v. Aguilar-Pena,
887 F.2d 347, 350 (1st Cir.1989). Thus, the Sentencing Commission has indicated that courts should consider departures only in “an atypical case, one to which a particular guideline linguistically applies but where conduct significantly differs from the norm”. U.S.S.G. Ch. 1, Pt. A, § 4(b). Yet if the disparate racial impact of the guideline ranges for “crack” offenses warranted a departure in
any
ease, it would warrant a departure in
every
case. One could hardly justify granting departures only for African-American defendants and not for defendants of other races, since the same offense would then lead to different punishments based solely on the defendant’s race. Even if this scheme were constitutional, it would violate the Sentencing Commission’s policy statement that race is “not relevant in the determination of a sentence”. See U.S.S.G. § 5H1.10. Thus, though
Cyrus
establishes that the relatively severe punishments for crack offenses serve rational policy goals and reflect legitimate differences between crack and powder cocaine, the courts would effectively have reduced the applicable guideline ranges across the board, using a power devised for atypical cases to cut the sentence for typical ones. We join several of our sister circuits in rejecting Thompson’s argument. See
United States v. Bynum,
3 F.3d 769, 774-75 (4th Cir.1993);
United States v. Haynes,
985 F.2d 65, 70 (2d Cir.1993);
United States v. Pickett,
941 F.2d 411, 417-18 (6th Cir.1991).
C.
Remaining Issues
Thompson’s two remaining arguments are foreclosed by circuit law. The unique ability of the U.S. Attorney for the District of Columbia to determine whether to launch a prosecution in federal or in local court is not grounds for a downward departure.
United States v. Clark,
8 F.3d 839, 842-43 (D.C.Cir.1993). And the “career offender” provisions of the sentencing guidelines, as applied to Thompson, neither constitute cruel and unusual punishment nor violate his due-process rights.
United States v. Spencer,
25 F.3d 1105, 1110-12 (D.C.Cir.1994).
For the reasons discussed above, we affirm Thompson’s conviction but remand his case for further consideration of his sentence. If the district court on remand finds a “reasonable probability” that Thompson, if properly advised, would have acted in such a way as to qualify for a 3-point reduction under U.S.S.G. § 3E1.1, then it should sentence him as if his offense level were 34 instead of 35.
So ordered.