Modeling Slopped Roofs and Impacts on Wind Design

Lately I've been running into issues with my lateral analysis on a large, square steel framed building. The roof is an 1-1/2" metal deck that slopes in one direction and drops 10' over the length of the building, which I've modeled as a semirigid diaphragm. No errors or warnings appear during the analysis but I notice some extreme shear magnitudes that are in the thousands of kips. These particular values only occur in the 'Y' Wind Load cases and only in the Fx component. 

I reset my roof slope in the modeler to a flat surface to the respective level and the extreme values disappeared, or rather became more reasonable and comparable to the other direction. Any insight for why this phenomenon is occurring or suggestions of better modeling methods would be appreciated, thanks.