It's not trivial to find that in the data. The cops are not required to keep the camera on all the time, or at least there is no consequence not doing so. They got the number from "administrative data", basically just police reports and civilian complaints, etc. Nothing related to the camera data at all.

Or it could also mean that, camera or not, the cop could get away with anything as if the camera wasn't present.

There are many other possible interpretations of the data, besides the one you provide above.

E.g. the cops can do lots of shady stuff that isn't a "use of force incident."

Yes, but that'd show up in the data. If there were complaints against them, and those happened to coincide with the cameras being off, that's a very easy thing to detect in the data.

As far as I can tell, there is no analysis of the video data mentioned in this article. They're only comparing numbers of complaints with and without cameras.