Arrgh. I still think you're fundamentally missing my point. You can't treat "black" via subdivision. The problem is mixing (i.e. most "blacks" are "half white", etc...), not imprecision. Basically you can't treat "black" at all. The term will never be useful to scientific study. The fact that there exist some identifiable ancestral groups will never (!) tell you anything useful about a typical "black" woman in america.

The same is true of "polynesian" -- sure, they might have been isolated at one point but by now almost every "polynesian" you meet in Hawaii has a ton of white and japanese ancestry too (substitute appropriate mixing for Maori or Samoan and Fijian, etc...), so what use is it to talk about the potential IQ effects his great-great-grandparents might have had?

Yet the argument at hand (that, as far as I can tell, you are in support of) is that somehow the "race" of real people can, because it is "genetically determined", be correlated with something like IQ, which it just can't. It's far too polluted a data set.

You're falling into the fallacy of the excluded middle. Just because "black" doesn't tell you everything about a person doesn't mean it can't tell you something useful.

That typical (now there's an overgeneral term!) black woman will have a darker skin color than a typical white woman. She will be less susceptible to sunburn. She is more likely to suffer from certain genetic diseases like sickle cell anemia.

Now, none of these are absolute statements. They are all probabilities. That does not make them useless. If someone comes into the ER with severe pain in their extremities, it is very useful to know that they're black. They may or may not have sickle cell anemia - that can be determined conclusively through a blood test and thorough medical examination. But knowing whether they're likely to is very useful information, because it lets you determine whether it's worth putting in the extra effort to diagnose it conclusively. (Similarly, even if they carry the genetic marker, they might not be having a sickle-cell crisis, and it could be a blood clot or some other medical condition. But it's pretty damn likely.)

It's a fallacy to believe that just because the data you're working with can't tell you everything, it tells you nothing. Rather, you should recognize the limitations of what you know, and use them to determine what else you need to know. Race tells you something. It doesn't tell you a whole lot, because there's a lot of individual variation within a race. But that doesn't mean it tells you nothing, either.