[danShumway]

You could very reasonably use a different denominator, but in that case the study would be measuring different things.

"Black people are X percent more/less likely to be killed than white people by police", is a different statement than, "Black people are X percent more/less likely to be killed than white people by police during a police interaction."

The first statement just draws attention to a very straightforwardly observable data trend, the second tries to figure out why the trend exists. It posits that the reason black people are killed at a higher rate might be because they have disproportionately higher numbers of police encounters.

And from there, people can ask why that trend exists, and so on, and so on.