All of this points out a lot of the problems with epidemiological data... averages can't be construed to mean anything to anyone individually. They often can't be used to indicate anything at all.
It used to be that the statistics were used to say that if you gave 10,000 people a dose of 1 REM, 3 people would develop cancer because of that exposure. The data was averaged from Hiroshima data and extrapolated down using the linear non-threshold model, which is losing credit. That was also taken to mean that if you gave one person 10 REM he had a 0.003% chance of developing cancer due to that exposure... statistically speaking. But is that kind of statistical extrapolation valid in calculating radiation exposure risk? We just don't know. None of the statistics are proof of anything and there is a good chance that the inferences taken from large doses don't mean a thing in small doses. Unfortunately it is the only valid data we have at this point that makes a lot of sense. The rest of the data conflicts itself.
As Marlin stated, we should be seeing far more cancers among nuclear workers than we are... we are actually seeing fewer than expected in a normal population. Either the numbers are screwed up, the statistics are being misapplied, the theories are wrong, hormesis is a fact, we don't have enough information yet, or we just don't understand the workings of the human body and its reaction to radiation exposure well enough. Take your pick, there is probably some truth in all of them... or maybe none at all and something else is at work. From a scientific study standpoint, we have a poor environment in which to conduct experiments, no 'blind' group to work with and a contaminated laboratory. Therefore all of the data is suspect.
The bottom line is we don't yet know what really causes cancer and what doesn't, why some people get it and some don't given similar circumstances. There is a theory that we develop and get rid of cancer several times a day, and only a failure of the mechanism designed to get rid of it produces what we consider to be the disease.