@dalek:
The problem here is we are all making educated guesses. I don't think there is anyone posting on this site who knows the exact logic of the feeds to the TANS on this particular aircraft.
Agreed, which is why my view is that it needs further investigation.
@AA:
If that algorithm is weighted to A then the difference between A and B is almost certainly significantly greater than 330m.
'If' I am correct then the 330m is the difference between C and D. With your information, as a flight continues A and B diverge. C will lie between A and B, but closer to A. D will be within the GPS error of A but not necessarily between A and B. As A and B diverge C will get further away from A and may get further away from D.
EG