Yes, you'd need the light bounce direction for radiosity normalmapping. This is currently not stored in the WMB file, so this information gets lost. It would make sense to store the directional components of the lightmap in a future version. I'll check if this can be done within the current format.