Statistical power.
Given that the Conc and AF447 represent 66% of the sample since 2000, it is hard to avoid the conclusion that any estimate of the baseline rate of accidents based on a total of ONE data point is predestined to fail.
Further, if French culture is inimical to working with complex technology, how does the world's biggest nuclear power industry manage? Seems to work better than that in the UK, US, Russia, or indeed anywhere else.