The only eyeopening part would be the fact, that Apple doesn't store much about it's users, which isn't verifiable.
Also unique identifiers are not the only way to link data. And due to differential privacy the guarantees that the data cannot be linked decays over time (see https://news.ycombinator.com/item?id=15224312)
Also, why would they risk lying? What's in it for them?
Depending on your threat model in regards to your privacy the assurances made by Apple may be enough.
To me, it doesn't really matter. As long as I have no proof or control over how my data is being processed, it's better to just assume the worst case and practice data minimisation.
That's neither obvious nor true. Absence of evidence is more likely if there really is an absence of whatever; thus it is in fact evidence of absence, if only weak evidence.