Genetic diversity within continental races, including that of Sub-Saharan Africans, are mostly a consequence of genetic drift.
While genetic diversity between races are from selection. Thus the inter-racial genetic differences are more likely to manifest in trait differences that humans find more meaningful (which I use purely in a descriptive manner, not prescriptive), such as physiological (medical, metabolic), psychological & behavioral (personality), cognitive (intelligence), and of course physical (appearance, athletic).
The intra-racial differences that arise from genetic drift result in things that are still tangible genetic differences, e.g. ABO blood group frequencies, but don't map well onto characteristics that human societies place emphasis on as much.
And to address your point that:
>The genetic diversity of "black" alone exceeds the rest of the world combined.
This is because the level of genetic diversity as influenced by genetic drift is primarily a function of population size, and Africa being the origin of the Homo sapien species, and probably the Homo genus as a whole, has always had the highest level of effective population size. Thus genetic drift in Africans is least likely to be able to cause allele fixation on particular genes, and so such diversity is better preserved. But as already mentioned, these forms of genetic diversity is less likely to impact the observed traits that most humans, both academics/social scientists and your average joe, find "meaningful".
OK, then let's do it right. But I think you know that isn't really the issue here. Nobody is putting out studies on correctly defined races by genetic groupings and intelligence either because the topic is still considered heresy. Your point that the commonly used definition of race is inaccurate is simply deflecting from this fact.
The point is there is no such thing as a “correct” grouping. The choice of what constitutes a group is completely subjective, it all depends on how far you choose to zoom in or out.
There are correct ways to group but there is no correct group size. Clustering algorithms inherently require some choice of sensitivity. There is no single correct choice for a given set of data, it all depends on what you’re trying to accomplish.
Many sub-Saharan African populations, such as Bantu-speaking West Africans, exhibit relatively lower genetic diversity compared to the Khoisan people, who typically have light brown skin. The Khoisan lineages diverged from those leading to Bantu and other sub-Saharan groups around 100,000–150,000 years ago, making them one of the most ancient human ancestries.
Most domesticated animal breeds are highly inbred so have much tighter genetic boundaries. It's much more similar to doing genetic research on human families, which is definitely fruitful.
Or more accurately, if it were genetic the races would look very different.
The genetic diversity of "black" alone exceeds the rest of the world combined.
So you have two choices:
1. Everybody is black.
2. The other races roughly stand, but there are dozens of different black races.
Or you can be more accurate and say race is cultural.