I can see it if m is, say, 90% of the entire population. But when m is measured in thousands, how is the vastly larger death toll ever rationally preferable just because it's spread out a bit?
It's not even spread out that much. Worldwide, coal kills more people in two months than nuclear ever has, and that's including nuclear bombs.
Still, n * 1000 deaths (in n years) may rationally be preferable to m deaths once, even if n * 1000 >> m
Risk assessment is a funny thing. Economists and psychologists have done great experiments along these lines.