GA isn’t arena, but I think a comparison to arena is apt:
In arena, depending on the choices you make, you either rise or fall to the level at which you’re able to be competetive, and you reap the rewards for competing at that level. So if you invest heavily in a new meta right out of the gate, your squad improves and you get better rewards. If you cling to an old meta for too many months, you gradually sink in the ratings and get worse rewards. But the system is self-correcting: you get weaker opponents if you start losing, until you fall into a region of the curve where you’re competetive; and you get stronger opponents if you start winning, until you rise to a region of the curve where the competition is too strong.
What’s missing from the GA matchmaking is this self-correcting aspect. No matter how many times a GA event runs, the pairing algorithm won’t learn your true strength, because it’s based solely on GP. You’ll keep getting grouped with people who have demonstrably weaker and stronger rosters, who either can’t possibly beat you, or who you can’t possibly beat.
(band yes, over the course of a 3-round GA event, pairings will gradually become more competetive since records in THAT event will play a role, but at the start of every NEW GA event, the matchmaking will go back to being as random and uncompetitive as they have been to this point.)