That would be fun to research, but my guess: if you showed them the shape of various countries, more people could pick out the right shape. But that shape cuts off Canada and Mexico in ways that are essentially arbitrary, so if they're keying on those edges, or on the grid-of-states shapes, they could still struggle with a satellite photo.

My guess is that this number would be significantly better today because people interact a lot more with world maps when they accidentally zoom out on, e.g., Google Maps pages.