This banding lark is becoming a bit of a pain. I was banded due to the same process you describe. If banding is being applied before G.INP then the DLM algorithm is wrong it should be the other way round.
This smacks to me of poor coding on DLM with G.INP coming in at a later date and the DLM coder not programming G.INP to kick in earlier.
Only way was for mine to be DLM reset by a BTO engineer but I had an incorrect underground feed that was losing 13m downstream speed. An overhead dropwire fitted underground and not twisted pair.
I'm still waiting for BTO to put my drive back together after digging it up. Partially done today I might add.