Yup, that's always in the back of my mind when people talk about javascript-enabled user stats. Of course, it's perfectly possible that the OP IS recording accurate(-ish) JS usage stats, it's just more involved than other client stats that are readily provided to the server, so less likely. If 99.99% wasn't just a figure plucked out of the air, it sounds believable that it's being obtained 'correctly', since it's not 100% :)