It's probably worth tracking and considering as a metric for matchmaking, but a couple potential downsides could be lowering the player pool where other parameters like latency have to be expanded, it having the potential for reverse boosting/smurfing in order to grantee stomping worse players, and at least some people not finding it enjoyable depending on implementation/complaining about "forced" 50:50 winrates.[1][2] Not to say it's necessarily worse than the alternative, but there's at least some discontent with at least some of the implementations.

Overwatch for example doesn't seem to try and fill in say 6 players of a range of 2500-2600 ELO, but will pull something like 2 higher ELO player and 4 low ELO players, partially due to there only being so many higher than average players to matchmake.[3] "Good" players get "bad" players to bring down their winrate and "bad" players get "good" players to bring theirs up. This miss-match might contribute to why some people become so toxic, especially when having fun relies on your teammates so heavily and everyone wants to be the DPS.

Also personally I enjoy playing against better players (within reason) in order to improve my own gameplay via imitation.

[1]https://www.halowaypoint.com/en-us/forums/84ad72a8b518479785...

[2]https://www.reddit.com/r/blackopscoldwar/comments/ifwuy4/sbm...

[3]https://win.gg/news/jeff-kaplan-on-overwatch-players-matchma...