There is growing evidence that criminals DO, at least to a meaningful degree (hey, they're criminals - most are stupid), choose their targets based on odds of the would-be victim fighting back. Of late and in particular, several mass murders have occurred in "gun-free zones" when more convenient/suitable (to the killer) venues existed but allowed civilian weapon carry.

Parallel answer to your question is that the measured violent crime rates you reference are too broad, aggregating a few small high-violence areas with broad low-violence zones, creating misguided statistics.