We live in a world where people disregard qualifiers, so if you say "on tests of mathematical ability, men have higher variance in test scores, so the most talented mathematicians are disproportionately men" people will hear "men are better at math" and assume that average men are better than average women at math (this might also be true, but is not what you said).

I could make similar argument about a lot of things we do here, e.g., people hear "consequentialism" and think "the ends justify the means", that doesn't stop LW from promoting consequentialism.

So as a general rule, we pretend that there are no between-group differences because if we don't, people have a tendency to focus exclusively on between group differences and ignore within-group differences, which is worse.

Intentionally believing false things always carries a cost.

For example, suppose I want to hire the best mathematicians for a project, they'll likely be disproportionately White or Asian men. Someone who followed your advise looking at the mathematicians I hire would conclude that I was racist and sexist in my hiring and we live in a society where the courts might very well back them. Thus the only way for me to avoid being considered a racist and sexist is to intentionally fudge the numbers based on race and sex, which itself requires that I know the truth about racial and gender differences so I know which way to fudge.