The Zwift categories are drawn at arbitrary lines that stand for nothing in reality. Whichever way you skin them, there will be examples where certain weight and power combinations will do well, and others will not, and open up an easy method for sandbagging - managing your output to stay in a category lower than you should so that you can get good results.
It is very easy to stay in a lower category than you should if you avoid big 20 minute efforts. Prime example - ZRL race 3. Most of the ‘top’ B riders contested the sprints rather than the KOM - because an all out effort on the KOM would have got them upgraded.
The only way to fix it is grouping by results (rankings).
https://forums.zwift.com/t/rankings-based-matchmaking-get-rid-of-cats/519476