Because you still don't get it. If your chamber is 1.630 (hypothetically) and your fired cases (90% by your measurements) come out at 1.626 then your chamber is at least 1.626 long, possibly 1.630.
The farther away from that chambers shoulder you get, the less chance of making contact. A case shorter than the chamber CAN'T be bumped by a chamber shoulder it never touches.
I'm not even sure why we're still discussing this. If 90% of your cases come out at 1.626 and you already bump to 1.623 with 100% reliability the discussion is over.
You want to know if it's worth not bumping 90% of cases that measure 1.626 because less than 10% come out at 1.628 and 1.629? The answer is no, it's not worth risking feeding problems with 90% of your cases to avoid overworking less than 10% by an additional .003
So 10% measure longer? Who cares? If the die is set to bump everything .003 under the 1.626" 90% of your cases measure on ejection, those 10% will get sized a little extra, and they'll all run 100% reliably in the gun.
Why risk malfunctions 90% of the time to benefit 10% of your cases?
Your chamber isn't growing longer. Perhaps the brass is springing back less, who knows. If sizing .003 under whatever the majority of ejected cases measure (1.626 right?) runs reliably why would you change anything?