DanJon,

Like you did, changing sign of U1 is not a problem, but are Win1 and Win4 has same value and direction(Check the direction cosine and parameters)? Changing sign is not a problem.

Load cases are equations to get the results you want.
You can make your pattern simply adjust load cases.

You can compromise its order, but it does have to make sense.
Assume simple load cases below
eg)
L1:W+P1+T1
L2:W+T1
L3:L1-L2

from above, you can get displ, and loads at the SUS, OPE cases from L1, L2 and EXP stress range from L3, but if we change the order like below,

L1:W+P1+T1
L2:L1-L3
L3:W+T1

you'll get the wrong values(Or not even able to precede), because CAESAR II not yet calculate the L3.

Well, I only use B31.1 and it syas we don't consider seismic and wind occured at the same time, so I'm not familiar with your load case above. But I guess the logic is same.