From what I heard the concern at Fukushima was that the earthquake had compromised the integrity of the containment structure, so allowing the core to melt down was judged as too risky (I'm not sure if I agree with that decision, but of course my opinion in hindsight is not relevant). In fact the majority of radioactive material that was released in the incident might have leaked from the spent fuel storage, which was also damaged in the earthquake:
http://web.mit.edu/nse/pdf/news/2011/Fukushima_Lessons_Learn...
You could have a much safer design available, but if the 'less safe' design is more cost effective, and meets all specifications, it would still be the preferred design. This means that the any problem is with the specifications/requirements. The next question is whether the standard of safety which would require choosing the 'failsafe' (or at least failsafer) design would cause statistical murder.
This is the one simple problem.
There are now reactor designs that are passively fail-safe. So that even if all the pumps and everything stop, it will still not result in an explosion or otherwise serious release of radioactive material.
I don't know how much those sorts of designs were developed at the time Fukushima was commissioned though.