We're reading the study quite differently. I think it's quite clear that the researchers were interested in neither legal culpability nor determining which party made the "worse" error in judgement. They quite clearly state that listing an accident as caused by one party or another is for classification purposes only.
I agree that we are reading it differently, and I admit your interpretation of the researchers' interests is a possible one - but it is far from clear. It certainly isn't clearly stated that the labels are for "classification purposes only" anywhere. Do you mean this quote?:
The categories are defined by the driver- or cyclist-actions that describe each event most succinctly. Categories named "Ride Out..." refer to the actions of a cyclist, while "Drive Out..." refers to the actions of a motorist. Although they may refer to the actions of only one party, these labels are not intended to assign fault.
Again the "assign fault" in the above quote means legal culpability. Possibly it means physical "cause" as well, but this is doubtful, as they are quite clear about making a specific distinction between "cause" and legal "fault" later on (when discussing cyclists riding on the sidewalk):
...While cyclists in such cases may therefore be deemed at fault in law, it is probably not correct to suggest that sidewalk cycling was the sole "cause" of these collisions. In many collisions of this type, it is quite likely that the motorist did not come to a complete stop before crossing the stop bar....
As you can see here, they are saying that while a cyclist riding on a sidewalk who gets struck by a car at an intersection might be found legally culpable they attribute the actual cause of the accident to the motorist's behaviour
Nowhere do they say the bins are arbitrary classifications. The bit about "...that describe each event most succinctly" indicates that the bins ARE related to the behaviours which precipitated the collision. Furthermore, their bin types are derived from (and meant to be comparable to) those of the Federal Highway Administration's (which you can find here: http://www.fhwa.dot.gov/publications/research/safety/pedbike/96104/). The FHA's crash types are specifically defined as:
These crashes can be classified or "typed" by their precipitating actions, predisposing factors, and characteristic populations and/or location that can be targeted for intervention.
So, the type classifications they're using are very much meant to indicate which behaviour precipitated the collision (and who's behaviour needs to be corrected)
In the quotes I included above the researchers disclaim both fault and absolute cause.
I believe I have already adequately addressed your first quote, but your second quote: "Thus it cannot be said, for instance, that more cyclists than motorists caused collisions by disobeying traffic control." is taken entirely out of context. It's from a section dealing with other secondary contributing factors that may have played a role in addition to the main cause. The sentence you quoted is referring to their inability to make an accurate comparison of cyclist caused collisions due to disobeying traffic control vs motorist caused collisions due to disobeying traffic control - because secondary contributing factors were not recorded in all cases. Here's the same quote with the relevant preceding context:
...Note that, while some of these factors (weather conditions, etc.) are known in almost all cases, others (such as disobeying traffic control) appear to have been reported less consistently. Thus it cannot be said, for instance, that more cyclists than motorists caused collisions by disobeying traffic control.
It's quite clear from that section that what they're saying is that of the collisions caused by motorists and those caused by cyclists, they cannot compare the proportion of each that were due to disobeying traffic control (e.g. running a red). So they can't say something like "Of the collisions caused by cyclists 60% were due to disobeying traffic signals, while of the collisions caused by motorists only 40% were due to disobeying traffic signals.". This necessarily implies that the crash types assigned do indicate an attribution of cause to either the cyclist or the motorist in the first place.
As for the rest of your last comment, you do realize that it is simply your opinion and perspective right? I could tell you a whole slew of similar stories about the poor behaviour of motorists as well, but it would only be my opinion and perspective. The point is that we are BOTH biased in our perceptions and observations, we BOTH only notice and remember the behaviours of "the Other" that piss us off, and we BOTH empathize more with the side that is most similar to us. So again, while I could go on with stories that are basically the same as yours but where its motorists driving recklessly, I realize that I am likely highly biased in my perceptions and memories. They are likely not representative of actuality. There's simply no way to tell from anecdotes who's correct, my opinion isn't any more valid than yours, and yours isn't any more valid than mine. The only way to really grasp what's happening is with empirical data.
While there is some small ambiguity in the interpretation of the study I linked, I also mentioned earlier that the same basic findings had been replicated in other studies. Someone else was nice enough to post links to articles summarizing some of those other studies elsewhere in the comments. I'll reproduce them here:
While it may seem to you like cyclists behave in a more reckless and dangerous manner, the data simply doesn't support that view. In the end though, it's up to you whether to ignore the data or re-evaluate your preconceptions.