Oh, you are absolutely right that racial groups are very crude and shouldn't be used when a better metric is available. But they are already used when assessing policing, education, hiring, housing.

I'm not advocating for writing race back into public policies. As you point out, it would be very unfair to some people. But naturally occurring group differences should be taken into consideration, and not as an evidence of unfairness. For example, racial profiling by police isn't bad overall. And schools would end up (mostly) racially segregated if you fairly segregate by ability. Some welfare policies lower selection pressure (all the way into the negative) and exacerbate the problems. A lot of corporate wokeness was initially driven by the legal threats.

I think within those groups there are subgroups that can vary within each other, probably a lot.

Of course. But the clustering refutes the leftist narrative that "races are social constructs". It shows that the stereotypical groups align with the best fit genetical grouping and geography. And they align much better than expected.

couldn't get the peer reviewed(?) study by Lynn

https://sci-hub.se/10.1016/j.intell.2014.08.004 direct link

There are better clusters, where Middle East has distinct composition even at K=2. K2-7 and subsets, K7. But the IQ difference with Europe could be partially environmental.

Notice, that 1 SD difference in IQ (85 vs 100) is within the US. Sub-Saharan Africans score below 70. And 2nd generation immigrants to developed countries gain IQ.