Well off the top of my head it'd be assuming that searches were distributed randomly w.r.t bad data, whereas in all likelihood there may be some (inverse) relationship between the badness of an area and its search popularity. In general living outside the US I was happy with the gmaps solution generally, apple maps had some glaring errors in my neck of the woods on launch.
Oh, please! There's no way to get the data required to do any real statistics work on this. Not without doing a hell of a lot more work than is worth for an HN post.
My simplistic 5% = 100 million calc serves to simply highlight that small numbers can mean huge problems. Even if I am off by an order of magnitude this is still a huge number of searches per month that are potentially sending people to entirely the wrong place, if at all.
Please go back to stats class.