95% of bicycle deaths are a result of motor-vehicle bike collision, and the vast majority of collisions are between the front of the motor vehicle and a bicycle.[0]
> there will still be bike-bike collisions and other types of accidents
Sure, but these collisions are far less deadly and injurious. Exchanging car-bike collisions for bike-bike collisions is a good tradeoff from a public health perspective.
[0]: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/...
I can’t find any hard data on this, but from what I can see online (and my own experience riding in the road), the most common types of accidents are when cars are turning, and fail to yield to a cyclist. Being rear ended or sideswiped seems to be much more rare than accidents where the car is turning and doesn’t see the cyclist.
For example, a cyclist is riding in a bike lane, and a car is turning left into a driveway. The car is looking for oncoming traffic in the road, so the cyclist is most visible in the road or bike lane, where the driver is expecting fast moving traffic to be. If we build a segregated bike path, cyclists will be much less visible, and it’s more likely that a driver will fail to see them when turning.