I agree that census data on the subject will be used by reactionaries as a basis to react. Or more generally, to polarise in any direction. I don't think that's the intended function of a census though, and I think mitigating that unintended outcome isn't best achieved by compromising the intended function. Function which I think is a net positive for society in an analytical sense.

e.g. in sports - if there was a legitimate cause for concern about a perceived unfair advantage to women that were assigned male at birth, wouldn't it follow that we would see a statistical overrepresentation of trans women athletes relative to trans women in the general population, vs cis female athletes relative to cis females?

To my knowledge what's been observed elsewhere is the opposite, supporting that there is no legitimate cause for concern. Not that there would be if trans female athletes were indeed overrepresented, but I think the reactionary argument falls apart for the majority where the data supports the opposing view.

It also occurs to me that limiting the data via either omission or underreporting could equally be used by those with malintent, e.g "why spend time worrying about the needs of 0.1% of the population?"

It's not really for me to comment on whether it's a net positive for trans people specifically, but in lieu of other info I'd defer to orgs like Equality Australia on that, who seem to think it's warranted.