Highway safety statistics have significant limitations with respect to fine grain detail and data collection capability. Any datasource that uses police data must account for jurisdictional differences in data collection as well.

Nothing is wrong with the statistics, it's how you use them. It's just difficult to meaningfully say that it's safer to take the bus from point A to point B using statistics that are generalized over time and geography.

For example, 30% of motor vehicle fatalities are alcohol related, and the vast majority of those fatalities take place after most people are done commuting for the day. Uber probably has a significant impact on preserving human life, after dark.

The purpose of the anecdote is to illustrate how those stats relate to decision making and risk/reward management by the travelling public. Am I at more risk driving three exits down the highway vs. crossing a busy avenue, sitting at a bus stop, riding a bus, crossing three other busy streets? I don't know -- and those statistics that you are citing don't answer that question. My guess is that I'm safer in my late-model car in a typical automobile accident than in any traffic or crime incident in the bus case.

They still have an important point about what statistics to use. Total death count is useless. You want numbers for per-mile and per-hour, and you want to look at more than just deaths.