Question:
Each year there are about 500 fatalities on bicycles related to head injuries. Only about 15% or 75 of these are single fatal injury accidents. Additionally there are about 1,500 serious head injuries to bicyclists almost all of which resolve to the victim returning to normal or near normal activity.

There are about 50,000 emergency rooms in the nation.

This means that on the average there are 1575/50000 = .03 victims per emergency room per year. Now certainly these accidents are not spread all around the US evenly and metropolitan areas probably get a majority of the accidents. But there are still over 500 metropolitan areas in the USA and this still comes down to one serious seriously head injured bicyclist per year. And during this time there will be 50 falling accident head injuries and 49 automobile accident head injury victims. -- Can you suggest why an individual doctor would ever have enough information to make a statement like the one that Dr. Kelly made? The facts are, to quote a Bell Sports PR representative (Paul Thatcher), "It helps to have safety experts say the things that would look callously self serving for Bell Sports to say." And in this vein Bell Sports has shopped around and bought research that invariably shows helmets with large positive results in sabving lives while virtually all of the privately funded research has shown not a detectable change in either serious injuries or fatalities attributable to helmets.

Answer: If you had actually read any of the reports that you listed citations for, you would realize that the original Thompson and Rivara study was based on a population of cyclists with abnormally high helmet use rates relative to the general cycling population at the time (over 20% vs less than 5%, ~1987 IIRC). These "early adaptors" of bicycle helmets almost certainly had quite different behavioral characteristics than the general population, being for the most part more cautious and law-abiding in their riding habits, more concerned with their own personal safety and well-being, and more likely to follow the vehicular cycling principles advocated by the LAB/LAW, CanBike, etc. As such, they would certainly be expected to have far fewer crashes of all types (both head injury and non-head injury) than less careful or more reckless cyclists. Remember, this was well before the safety zealots fastened onto their one-shot approach to cycling safety and started pushing their mandatory helmet law agenda with a vengenace. Once this happened and bike helmets were marketed ever more aggressively to the cycling masses largely ignorant of vehicular cycling practices, including children, risk compensation behavior presumably kicked in to negate or even reverse the expected safety benefits. Indeed, I would argue that by vastly overstating the theoretical safety benefits of helmet use, the net result was a serious loss in terms of overall safety.