If you look at the distribution of Japan's nuclear power plants, they actually tried to put them in the safest areas: http://www.oecd-nea.org/press/2011/NEWS-02.html
Where there are nuclear power plants along Japan's eastern coast (close to the subduction zones), they're mostly in the north.
Japan has two subduction zones. The Japan Trench in the north, where the Pacific plate subducts, and the Nankai Trough in the south where the Philippine Sea Plate subducts.
These two subduction zones are very different. They're on opposite ends of the spectrum in terms of earthquake behavior.
Before the Tohoku earthquake in 2011, we had never observed a magnitude 9 earthquake on a subduction zone like the Japan Trench in the north of Japan. (The Japan Trench usually has frequent magnitude 7's. Damaging but not devestating.) All of the previous magnitude 9's were on subduction zones similar to the Nankai Trough in southern Japan.
Furthermore, southern Japan has a 1300 year written historical record of large earthquakes and tsunamis occuring every ~200 years (they're not regular, but that's another story). The last two were in 1944 and 1946. In contrast, northern Japan has just as long of a record, but no records of very large earthquakes and tsunamis. (There's some sedimentological evidence of a very large tsunami in northern Japan ~2000 years ago, but prior to this, we assumed that it must have originated somewhere else.)
Therefore, southern Japan has the highest hazard for large tsunamis, and northern Japan has a much lower risk. Northern Japan's risk is now higher than we thought it was, but southern Japan is still more likely to have the next big tsunami.
Notice that there are very few nuclear power plans on the east coast of southern Japan. Instead they're on the west coast, away from the main tsunami hazard. That's why.
At any rate, the Tohoku earthquake really turned what we thought we know about very, very large earthquakes on its head.