[pcl]

My point is that the author is advocating collecting statistics in a way that equates gang shootouts and backfiring cars as equal incidents of gun violence.

You've identified a problem with using human reports as a means to measure the number of shots fired, whereas the author suggests that some abstract "shots fired" count would be a better measure than homicides and injuries.

The author doesn't propose a suggested methodology for how to do so, but the metrics that he presents are from ShotSpotter deployments. These sorts of systems are designed to identify gunfire in particular, and so are presumably more reliable than monitoring 911 traffic.

Note that he identifies a number of other weaknesses with ShotSpotter-based data -- doesn't pick up indoor shots very well, and is of limited nationwide coverage, for example.

More details: https://en.wikipedia.org/wiki/Gunfire_locator