This would only be true if affirmative action were carried to the point where the percentage of black students in the elite schools exceeded the percentage of blacks in the general population. I don't have the numbers handy, but I did go to grad school at an Ivy, not terribly long ago, and that does not match my recollection of the racial make-up there. The undergraduate ranks seemed to be dominated by rich white kids.

Even if that were true, affirmative action is based on admitting a certain percentage of blacks.

This sort of quota approach isn't the only kind of affirmative action, although it's possible it's the only kind implemented in universities?

Unless I miss your point, this only holds true if the percentage of blacks required to be admitted is higher than the percentage of blacks in the population.

One reason for thinking that a measure of talent is poor might be that it is outperformed by other measures. There may not be genuinely good measures of talent. It does occur some sort of retrospective measure based on results is probably better than what the admissions office uses, but that is surely still not a perfect measure, and is also obviously not a practical option to replace what the admissions office uses (unless someone invents a time machine). Another reason to think a measure of talent is poor, though, and this is probably more applicable here, is that a measure may be considered suspect if there is reason to think it is really measuring something else entirely, perhaps because it correlates suspiciously strongly with factors regarded as independent of talent.

Those usually considered by admissions offices are known to be horrible and in fact were originally selected so as to allow tacit discrimination while maintaining a veneer of fairness (specifically to discriminate against Jews, who by previous measures of achievement would have dominated the Ivies for a couple decades. Asians are currently in that same position.)