But the study is tiny, was designed to measure something else, is in only one country, and can't even statistically distinguish that the "early period" is different from the "late period" because the error bars are so wide!

You don't convincingly demonstrate to doubters--or to yourself!--that anything is the case with such poor data. It's moderately suggestive, and that's all.

If someone is trying to use that study to prove trans criminality, then we can confidently say that their case is very weak, but the support for your alternative perspective is also very weak. (It's literally a few dozen people total, spread over 30 years.)