While trying to collect unused bytes here and there in the LN2 disassembled source, I figured out an optimization that shortens the code and makes one of two lookup tables unnecessary. If I were able to consolidate all the unused fragments before $2000 (where the bitmap starts) I would probably be able to gain a 512 byte slot for some additional code. In there I would move a few routines so that I could extend the sprite clipping routine to work 100% of the times.
The sprite clipping problem is quite complex to explain so I will try to put together a few images of where the tricky bit is and why John Twiddy’s simplification saved him big time.
As I commented before, this problem seems to have been solved properly in LN3 (well done Stan!).