I don't think the bimodality argument helps their cause. With respect to gender (not sex), bimodality is a fact: if you were to measure a number of characteristics that help to define what we think of as gender along a continuum, the multivariate distribution will in fact be bimodal. What that means is that there are two peaks, where most people are. One peak is "male" and the other is "female". Yes there are some people in the valley between the peaks but -- and this is the important point -- by definition of the term "bimodal distribution" there are relatively few people in that valley. It reinforces the point that continuous "gender" characteristics, which individually do lie along a spectrum, collectively organize around two poles that (shocker) are closely linked to binary biological sex. The variations correspond to GNC behavior, which should be recognized as variations along some axes but not wholesale flipping (except in a few cases), and in no way corresponds to a change in the biological ground truth of sex.