Listening to your track again just now: you might be able to pull it off with just mercyless application of EQ - basically modifying the sounds of the individual voices by almost cutting out areas in the spectrum, that the other voice is untouched in (or even boosted) to clearly distinguish the lines from each other.