How can that number possibly be accurate when 1) gender and sex are these spectrum/grid things, and 2) most of us are in the closet anyway?
I don't know that gender and sex are a spectrum
for kids. Rather, I have the hunch that for most kids, especially very young ones, being trans is a straight-up, very traditionally binary, girl/boy thing.
My reasoning for this is that a more nuanced concept of gender is something that a) is rare anyway, and b) takes considerable time to arrive at.
Simpler gender systems (such as a strict binary, or a binary plus a "third gender") are much easier for younger kids to understand.
I mean, it's so much easier to say "I'm a girl" when you're only four years old than "I'm a poly-gendered person with a preference for feminine attire and the colour pink and my preferred pronouns are xe, xem and xir".
Maybe the observation that most trans kids change their minds reflects this evolving concept of gender. While early on a simplistic view of gender might make you think that "I'm not a girl" because you like to play with fire trucks rather than dolls, once you've grown up a bit you realise that girls can dig fire trucks just as bad as boys do, so maybe wanting to play with fire trucks doesn't make you a boy.
Also: kids' identities are very flexible and it's a mistake to consider them from the point of view of rigid adult identities who have had a lot of time to settle and much opportunity for refinement.