The risks minimized for nuclear are only those well-known and quantified beforehand, and not even all of them (those that would make nuclear much more expensive, like retrofitting protection hulls, are ignored). Yet several incidents show up risks that weren't realised by the risk calculators. The appearance of such is more likely in more complex systems. Thus your risk calculations cover always only a fraction of the actual risks.

_{*Lunatic*, n.
One whose delusions are out of fashion.}