But, again, what about me here? The results he's getting line up pretty well with my experience. Why does your anecdotal evidence trump mine?
Quite apart from the weight of the comments in this thread suggesting
your experience is the anomaly (and do you
really see 95% of clerics multiclassed?!), it's quite clear that the research methodology is flawed.
1) In my experience, the vast majority of player characters used in actual play are not created or levelled on D&D Beyond. That's anecdotal, but in order for the research to be valid it's the responsibility of the researcher to demonstrate that at least a significantly large sample of player characters used in actual play
are created on D&D Beyond for the results to be valid. They are also only considering people who have unlocked all options. If you take a small sample size and make it even smaller you are going to get all sorts of weird anomalies.
2) We have very good reason to suppose that the vast majority of characters that are created on D&D Beyond are never used in actual play. The authors' claim to have accounted for that, but they have not demonstrated that their selection process
actually works. In order to do so you would need to take a sample of the characters your method has produced and survey their creators to find out if they are actually played.
3) It assumes independence: i.e. what character someone chooses to play in a party is not influenced by what the other players choose. This is quite clearly an invalid assumption. If a party already contains two of any character type it is unlikely anyone will choose to play as a third.