The first thing that comes to mind is that your system may be undergoing non-linear restraint changes between Sustained and Operating. Perhaps the wind load is causing a pivoting in the model that causes lift-off from a support?

(The other thing would be some type of input error, but I presume you have checked that out?)

Note that if you have a non-linear system, running a load case as W+P1+Wind1 (OCC) doesn't really satisfy the intent of the code. Your load cases need to be something like this:

1) W+T1+P1 (OPE) := standard operating case
2) W+T1+P1+WIND1 (OPE) := operating plus wind
3) W+P1 (SUS) := standard sustained case
4) L1-L3 (EXP) := standard expansion case
5) L2-L3 (OCC) := operating difference, gives wind effects
6) L5+L3 (OCC) := stress summation, SUS + OCC

Cases 3, 4, and 6 are your code compliance cases.