If you have a system with no gapped supports can U1, U2 and WINDX etc be run as seperate load cases in OCC and then combined with deadweight to get occasional stress? Or should they be combined with deadweight in OPE cases first and then subtract deadweight out to get more accurate results?

Make sure you have no other non-linear boundary conditions. This would include: friction, directional restraints (like +Ys), or large rotation rods.

I don't recommend doing things this way, you only save yourself one load case. You run the rist of getting into a bad habit, and worse, what happens if the model changes at some point in the future and non-linearities are introduced?