No idea about SF, but where I live, cycle-related facilities are split into 20% that are clearly the cyclist's fault, another 20% that are clearly the driver's, and 60% that either involve a combination of errors or where the details are difficult to ascertain, particularly since one person involved is dead.

If you assume that the 20% of fatal accidents that are purely the driver's fault are evenly distributed among cyclists, can you also assume that any 10-20% of cyclists are involved in 90% of accidents?

Fatal accidents don't have to be like general accidents. But.