That sounds highly implausible. I very much doubt that they went into the field and observed numerous police interactions and recorded them. If so, there should be also data showing how much their collected data agrees with what the police collected.

I think what you mean to say is that researchers analyzed the data rather than the police. However is all likelihood the data fed in to the statistical analysis was recorded by the police. Hence that data has a heavy bias.

Regarding better data. That is hard to obtain since it is the police which is in charge of recording it. The best one can do is using natural experiments create statistical analysis of bias in police reporting.

I cover some examples here: https://link.medium.com/fbjArValkjb