And the 'most people survive' is something of a misnomer.
So, if we consider only the accidents that are least likely to be survivable (which are a tiny percentage of all accidents), we find that they're not likely to be survivable?
Well, you could pull out two accidents out of the list where everyone died and say "most people die in plane crashes", which would be incorrect as well. Hence the term misnomer :-) As an engineer I would be inclined to say "if you find yourself in a survivable crash, the following things help increase your odds, ..." which keeps the cool bits about back of the plane vs first class, exits vs non-exits, but side-steps the probability question of surviving a gross malfunction in a commercial jet.
So, if we consider only the accidents that are least likely to be survivable (which are a tiny percentage of all accidents), we find that they're not likely to be survivable?