Thanks for the links. Particularly the first one makes for some very interesting reading. However, none of them were able to "break" or otherwise deanonymize any ChipMixer users, as far as I can see. The first paper specifically notes in its discussion that "Our approach only traces Bitcoins that are sent to mixing services", and that "Tracing Bitcoins beyond mixing services is much more complicated". Tracing bitcoins which are deposited to ChipMixer is not that difficult due to its unique chip payout structure, but if you can't link deposits with withdrawals, then your analysis completely breaks down and user privacy is completely protected.

As you say, outputs from ChipMixer which have been deanonymized have been so because of other factors, such as being combined with non-anonymized coins or deposited to accounts with linked personal information.