Any problems with the study? Control group is "random distribution of dogs" and the other group is "18,000 but of a no specific breed". Is that a problem or does it even out.
Importantly... the behavior is self-reported (by dog owners), who almost certainly have some bias about their dogs. I know I do. :)
Also what would count as "no issues" by and large depends on the dog breed I'd say. If a rottweiler tried to attack a rabbit that wouldn't be out of the ordinary, but if a border collie did the same it would.