I thought the same thing... It needs a finer scale, and maybe some adjustment for density of restaurants, since that probably influences the frequency of health inspector calls.

It looks like there's a nice gradient along Long Island, though. :)

The copy doesn't make it clear enough, but we're actually doing our best to adjust for small samples with a little empirical Bayesian magic. As the number of samples (inspections) goes down, the result is closer to the mean for the whole city.