Hmm. That is interesting to me.
It was not apparent to me that the base of the buildings as seen in the satellite photos have any more accurate coordinates than the rooftops. Michael's point is well taken in the case of skyscrapers but I was thinking more of buildings with very few stories.
Logic suggests that if the roads are accurately aligned then the bases of the buildings must be at their actual coordinates as well, but if the photo is not purely orthogonal to the current view (or at least very close to pure), then already the correctness of the coordinates of the entire scene is called into question, the further things are from the road. Yes, as the elevation of landmarks increases, the inaccuracy would be seen to multiply, but even at actual ground level things will have already begun to go wrong, since the satellite or plane was not pointing straight down, but rather slightly sideways, compressing distances between objects as you traverse the ground.