The problem is that total potential harm from a nuclear accident is very big. It is limited, but if the limit is all fission materials from the reactor disperse

The problem is that total potential harm from a nuclear accident is very big. It is limited, but if the limit is all fission materials from the reactor dispersed into the air or used to produce a real nuke by terrorists, then the cost can be loss of millions human lives. Compare that with aviation where the worst that can happen is a loss of airplane and few buildings. Given this potential threat anything that will be discovered to increase its probability will increase operation cost of reactors in the future. This is just not the case with planes. There minimizing the risk beyond a certain threshold is sufficient especially since it is known that an error in estimation of that threshold is bounded.