Yes, of course the variance of the distributions is greater than the mean difference. That is trivial. But it ignores the actual point. Mean population differences have practical population-level effects. The fact that internal variance exceeds the mean difference is inconsequential and meaningless.
It's literally textbook science. Like, if you pick up a university psychology textbook on sex and gender, you will probably find that statement in it.
It's based on evidence, not sentiment. That pattern is exactly what you see when you look at graphs showing the distribution of results broken down by gender on a whole bunch of physical and psychological tests.