Look up Banqiao at Wikipedia. The failure was only peripherally a consequence of the
dam and far more one of poor system planning, poor engineering, poor management, poor communication, a thousand-year storm (1,000+ mm rain in 24 hours), resulting comms failures, and the inevitable dam failure.
A relatively small proportion of the deaths were directly due to flooding. Most occurred from disease and starvation -- far more infrastructure and response failures than the catastrophe itself.
Even given massive flooding, access to a solidly-constructed building 2-3 stories tall would suffice to provide protection from immediate risks, as would comms, evacuation and response plans, emergency and relief supplies, etc.
China is a country which in the past century has seen tens of millions of deaths from starvation alone, most recently about 20 years before Banqiao. On the scale of Chinese tragedies, of which there've been many, this was relatively modest.
And, as noted, the region is more than fully recovered.
Contrast Chernobyl and Fukushima. 30 years after Chernobyl (the anniversary will be this April), primary containment still isn't achieved. Radiation risk was presented across an entire continen, including the 450 million inhabitants of Europe. Eating of wild foods in Germany is still restricted (particularly mushrooms and boar). Fukushima likewise still hasn't stabilised five years on, and remains uncontained.
Nuclear accidents have very long risk tails.
And all of the management, engineering, communications, and response failures of Banqiao apply equally to nuclear power.
The industry is quite fond of downplaying risks. Has done so since the first experiments. Wanting a thing doesn't make it so.