I have a better answer for you now.
For anti-fingerprinting you have 2 options:
A) Generalization
B) Randomization
Randomization means you (almost) every time have a new, unique fingerprint. This is not a problem, as it's new and thus you can't be identified.
Generalization mean you share your fingerprint with many others.
AFP usually used the randomization approach.
About unique fingerprint - even with Anti-Fingerprint feature the fingerprint will be unique (for example for device id) but it will still be fake each time website will ask for it.
So fingerprint is unique, but our user cannot be recognized by this website on next visit, and this is the purpose.
Anyway, we'll work on some general improvements in H2 this year.
I hope this answers your question.