Oh, that's interesting--for everything else I'm familiar with at least some literature on it, but I haven't come across anything that covers this. Do you have something in mind that you could share?

I agree that when carefully followed, the WPATH standards are pretty good. I don't think the Amsterdam gender identity clinic follows WPATH exactly, but the research from them seems to indicate that they're robustly identifying the underlying phenomenon, not being swept along by any sort of trend.

Regarding trans children historically, it's really hard to compare past and present incidence given changing societal expectations (e.g. if you have very very strong strictures against gender nonconforming behavior, noting it when it arises is more informative than when you don't police gendered behavior really at all). I'm not sure how to get anything like this solid enough to be worth doing math on.

Regarding detransitioning, the research is, frankly, pretty bad, because most studies fail to guard against selection effects. The problem for knowing about detransitioning rates is that the larger studies lose track of a lot of people, and assume but do not check that those who they lose track of show the same rates as those they keep track of. And the transgender survey is specifically targeted at people who identify as transgender, which again isn't necessarily as good at capturing people with a "this was a mistake" attitude and therefore probably skews the reasons you get. And the data might not be stationary (Amsterdam has checked and it seems pretty stable for them, but most studies don't check), and results probably vary by location. So, again, if the precise numbers matter, I don't think we have them. If we want to know if it's more like 1% or more like 50%, we're very sure it's more like 1%.

The problem with doing math on these things is that without reasonable error bounds on the numbers, you can't really tell whether the math is teaching you anything.