Personally I think they are missing one huge dividing line.

Cyclists are not a homogeneous group. Relative to this study it might be best to thinnkof 3 distinct groups.

Recreational cyclists.
Voluntary commuters
involuntary commuters

Each is different and there are secondayry differences.

The classic more cyclists less accidents was established with recreational cyclists. There is a very strong tendency for this class of cyclists to be in the same place and abotuthe same time. Thus drivers have recently seen other riders on hte same route and are aware of cyclists right then and there. Recreational cyclists are also more apt to have some training, perhaps not formal, but something picked up from others on the same path/trail.

Commuters are a different story. Generally there are not enough to establish a presence on the route. Thus drivers have not seen a lot of cyclists on their route, they are not any more aware of that there are cyclists then and there.

I personally think voluntary vrs. involuntary may also prove to be important. Voluntary commuters are more apt to actively seek some training, even if it is only asking on a board like this one. The involuntary ones may be in a position of having no choice. There may be enough people in the involuntary group who are there because of bad choices. It would not be surprising if they continue to make bad choices. (Not all by any means, but enough to explain things).

remember we have experienced an economic downturn, no surprise if there are more involuntary commuters. Also the involuntary commuters may actually have no choice. The other 2 groups whold then be more apt to avoid cycling under bad conditions.