The way they calculate is a bit odd
From what you describe, it is. A meaninful list would look for clusters of various crimes, and list the most saturated clusters as the most dangerous areas, whatever the population, zip code, or within one locality. Whatever the population is in those areas saturated with crime is doesn't really impact their danger. The population does not affect your safefy individually when you're in that area.
If I have a 50x50 mile area that has one murder, but a population only 200 people, it will indicate a very high murder rate per 1,000. But you have to ask, is there anything special about that area that made that one murder more likely? If not, its just an anomaly.
Also, if I have 5 apartment complexes, full of equally dangerous people, that share a common park, and most crime is in that park. Grouping one complex with the park will show it to be more dangerous, but it isn't. In fact, it may not even be the park that is dangerous. In the case of the area in Jacksonville, the murders involved people coming into the area to deal in crime and fight each other. You weren't likely to be threatened by that, if you weren't involved with it.