the on-chain footprint of coinjoins is much more obvious. blockchain analysis heuristics based purely on common output sizes are unlikely to be very useful.
Not only the output amount. How ChipMixer creates the chips follows certain patterns, you need to emulate the patterns too. Just splitting your coins into ChipMixer's chip sizes in your own wallet won't help against analysis like the FAQ claims. It can only fool stupid analysis tools.
you said "it's also easy to tell if you used chipmixer, because of the common chip sizes" so i was speaking to that, not about trying to mix coins in your own wallet. the point was that common outputs sizes are just that---common.
There are more problems. Many users sweep different chip sizes in single TX making it even more obvious. Even if none of these worked, then chips will lead to someone's tainted coins and will raise alarms at the exchange etc.
any examples of that actually happening?
i agree that many users sacrifice privacy by sweeping outputs together, and/or immediately, after mixing. ideally they would hold onto the private keys until they need them (throwing off time attacks) and spend them one at a time to avoid re-associating them.