Not following your point, unless it's merely to say that parsing observable reality isn't good enough to determine the precise levels of hormones flowing through someone's body, which, fine. That's certainly true. However ... physiological characteristics are in fact dictated to a large extent by hormones, and as a result one can realistically make a good inference about e.g. who has more testosterone.

my point is: biological sex follows a bimodal distribution with some in-between data points that are not trivially assignable to one or the other pole, called "intersex" humans. they are a biological, medical reality, and they themselves ask to leave children alone and allow them to decide when and if at all they want medical interventions.