The bed of my Taz5 is lower in the middle than the corners. The corners barely allow a business to slide between bed and nozzle… The middle is noticeably easier to pass the same business card. I just got a dial gauge to better document this.
An autoprobe only determines the plane of the bed. Then the firmware assumes the bed is flat. Therefore this type of warping / sag won’t be resolved with the autoprobe and three point sampling. Adding a few more sample points near the center of the bed may help “average” out the plane to compensate for the sag in the middle of the bed, but the bottom of a very large part would be slightly uneven. A raft could help, but my large prints often warp with a raft.
I’ve read that PEI has a natural curve when warmed depending on the thickness of the material. The TAZ5 PEI looks pretty thin, but as does the underlying glass… So this shouldn’t be a factor. I’ve gotten to a point where the ABS adhesion is good in the middle of the bed. I suppose, it could be a positive concept to vary the adhesion… Center less adhesion and outer edges better adhesion. I’m sure that wasn’t the intent, though.