the DL interference can be found easily by checking your frequency planning on a map.
the idle band measurements are only about the UL interference !
since you already changed the frequencies many times, I suggest the problem is not interference 🙂 when changing frequencies, do you see any change in the ratio of HO DL QUAL ?
Don’t you think it could be a Coverage issue ? Bad RXLEV in an urban area is usually linked to a bad RXQUAL. The HO QUAL might be triggered before the HO LEV. What is the ratio of HO LEV (UL & DL) ? What is the ratio of PBGT HO ?
What is your TCH allocation efficiency per TRX ?
What are the HO Success Rate per neighbour ? (type 180, per adjacency, not per cell)
What is the TCH Erlang per TRX ? Try increase the TRX PREF MARK (from 0 to 1 for example) on the TRX that carry the smallest amount of Erlangs, and oppositely, decrease the TRX PREF MARK of the TRX that carries the highest amount of TCH Erlang.