A quick note in regards to the above measurements, in particular the ANC off ones, I think that the differences below 1kHz and especially the notches can be explained by the 5128’s tendency to show “rocking” modes that may not occur in real human ear canals, combined with the ANC off Adaptive EQ function working in a different way from when the feedback loop is activated (all other modes), and not having the capacity to apply very fine grained filters to counteract these notches.
This measurement was done in a clone 711 coupler with putty to stabilise the Airpods to avoid any rocking mode and the difference is lesser.