Hypothetical: tutor group of 15. 8 boys, of which 2 are trans (female) & 7 girls, of which 2 are trans (male), and in which the females in this group are all struggling with their hygiene needs & the males aren't (not an entirely unrealistic scenario if there's an issue with break times - ref. theatre queues at intervals!)
Identify as girls / identify as boys question. Results show 75% of boys find break times sufficient for their needs & 29% of girls find break times sufficient for their needs. Conclusion: OK, this is interesting - let's investigate further, & we may take action accordingly (but we don't feel especially obliged to - phew!)
Male / female question. Results show 100% of boys find break times sufficient for their needs & 0% of girls find break times sufficient for their needs. Conclusion: Shit, this is clearly a real issue impacting female students because of their biology, & we must take action immediately (& we really can't avoid doing so - the parents would play hell after seeing these results!!!)
Different outcomes. Girls disadvantaged.
You can argue about degree: it's not by much, the trend is still clear etc. But why would anyone argue about degree? It's still a disadvantage, relative to what used to be. That's, arguably, ethically indefensible.
You can argue about my example: it uses specific, small numbers to make a point etc. But why would anyone argue about an example? There's still a possibility - probability - of data being corrupted to a greater or lesser degree, and that possibility didn't used to exist. That's, arguably, ethically indefensible.
Possible, very simple, solution: include the male/female and an optional gender identity question - all needs addressed, plus the additional benefit of possible insight into whether trans-identified females/males are struggling more (eg. due to toilet avoidance / stigma).
But the key point is that, in just this one example, in this survey, girls were redefined and they did did become less. As less is a relative term, there's simply no argument with this - their data and needs are being treated as less important than they were in recent years, and this inevitably means these needs become less visible.