Faking data is much harder: it’s easy to forget to remove something, change numbers so they don’t match other numbers, or make up numbers which have the statistical signal of manufactured data. Even being a great power does not prevent these kind of mistakes, because of the skill level censors would need in multiple scientific fields to make a good forgery. People get away with it in minor papers in the social sciences all the time, but with a million eyes on the data sets in this case, someone’s gonna see it.
They wouldn't have to actually fake scientific data - they'd just release everything else they were working on, omit the problematic files, and go "see? we showed you everything - there's nothing to see here!".
Time tables, budgets, staffing docs, those are readily forgeable. Any leftover discrepancies can be chocked up to clerical error. This is doubly so if the problematic research was classified or compartmentalized - files pertaining to those programs are often designed to be easily expunged.
Ultimately, China could release as many files as people ask, and never be able to settle the debate. You can't prove/disprove the existence of files they claim do not exist.
Our best bet would be whistleblowers - ideally multiple trusted and independent ones - who can testify to a hidden program. However, we can't use the lack of whistleblowers to prove the lack of a program, which is where we're stuck today.
