What work is the word "really" actually doing here?
How about referring to the cluster structure of gender space. Of course, then we'd reach the conclusion that there are only two genders, and the traditional assignment is people to them is the correct one.
"Once gendered behavior has been determined, however that occurs, cisgender males don't say "I'm a boy!" for cognitive reasons that are substantially different from the reasons that transgender males say "I'm a boy!"
Except "cisgender" boys don't generally engage in questioning "am I really a boy".
As I understand it, there is a phenomenon among transgender people where no matter what they do they can't help but ask themselves the question, "Am I really an [insert self-reported gender category here]?"
The obvious answer is "No". In fact this experience seems suspiciously like trying to make oneself belief that one believes one's gender to be X.
Humans universally make inferences about their typicality with respect to their self-reported gender. Check Google Scholar for 'self-perceived gender typicality' for further reading. So when I refer to a transman, by my model, I mean, "A human whose self-reporting algorithm returns the gender category 'male', but whose self-perceived gender typicality checker returns 'Highly atypical!'"
And the word 'human' at the beginning of that sentence is important. I do not mean "A human that is secretly, essentially a girl," or "A human that is secretly, essentially a boy,"; I just mean a human. I postulate that there are not boy typicality checkers and girl typicality checkers; there are typicality checkers that take an arbitrary gender category as input and return a measure of that human's self-perceived typicality with regard to the category.
While we're assigning categories in complete defiance to common sense and evidence, why are we so sure that the category "human" is applicable?
Yes, people tend to vigorously contest things that threaten their identity. This tends to have no relation to the truth of those things.