Assumptions are the bane of any statistician - they come back to bite you no matter how careful you are in allowing for, controlling for, or otherwise explaining why they might not be as big a concern as others may think. I think Bob has done an admirable job of justifying his assumptions, but they will never be accepted until proven properly. Which will never happen with the amount and quality of data available.
Another factor that must be accepted is the perspective of the researcher. An engineer will look for a design/maintenance problem. A psychologist will look for a human factors problem. An educator will look for a training problem. A DAME will look for a medical problem. A regulator will look for a blame problem.....(okay, maybe that's a little cynical). Put a team together of all of them (except maybe the last) and decide together what data needs to be collected by each to investigate the problems effectively, and go out and collect it for a decade. Then maybe there will be no need for assumptions. In the mean time, cover the assumptions as best can be done, and then do something. As sure as anything, with fatalities on the rise in recent years, if RA-Aus doesn't do something, the regulator will.