I think you missed the point of that XKCD. If you're going to publish a heatmap of absolute numbers, then yes, it'll end up looking pretty much like a population distribution map. If you publish a heatmap of percentages, then you will probably get something quite different (assuming that the percentage isn't linearly correlated with population).

This is often true, but until you see the heat map you don't know that. When its not true, things get interesting.

I think this case may not completely match the joke. In this case, I have a feeling due to issues of poverty and education that, for example, the rural South might have a higher arrest rate then the rural Midwest even at locales of similar population density.