Ok, I went back and started from scratch to make sure the whole process was controlled.
Within GIS, I loaded the terrain (DEM TIFF), generated a Contour (SHP) from the DEM file and loaded the photo mosaic (TIFF). I created a cropping mask and cropped a small region so I didn't have to deal with a large data set. Note that adding all these layers, which all use the same coordinate system, all line up fine within GIS. This is the cropped result within GIS. You have the photomosaic and the contours overlaid (the DEM is underneath and not visible):
Within VW, I created a blank plan, changed the units to meters and set the georeference origin. I then proceeded to load the contour shape file, which loaded where it should (0,0) and with the proper dimensions.
After that I proceeded to load an "image" (the photomosaic geotiff). The image loading dialog does not offer any option related to georeference. It doesn't ask to use the embedded coordinate system or assign something else as it does when loading a shape file. The image is loaded but its origin is not adjusted. It ends up thousands of miles away:
That makes sense. Provided it uses the embedded georeference to map it to the terrain. Is that documented anywhere?