How are you separating.
If you go composite then there will be an extra ink/plate generated called 256CVU, it would be up to your RIP to extract this. If you go separated quark would normally put the whole composite EPS in the code of each Separtion so again it wou require the RIP to separate properly. I'm guessing that the ink mapping from CVU to U that quark applies to the separated postscript is confusing the RIP.
It may be that you need a tool like PItStop to map the CVU to U in a composite pdf. Then Separate that.
... there was the puzzle of why the sun came out during the day, instead of at night when the light would come in useful.
Terry Pratchet