There is virtually no difference between using the OPA or not (I understand that I should not either, right? Opa here is not working as a servo). The four buffer have more or less the same offset (8mvolt)
All fets are matching around 530 ~ 540hfe and 540 ~ 550hfe (P to P and N to N) Is a closer match necessary?
If so, which would be the most critical?