I've been thinking that the success/failure discussion is the wrong way to be looking at this.
I remember reading (Rodney Turner?) that when PM is applied projects are X% cheaper, faster etc than if no project management is applied, but that the increased efficiency had to be measured against the increased costs (i.e. you don't put a PM onto a minor system enhancement.)
So, rather than talk about failures what we should be talking about is relative improvements.
(Dan Strayer was the catalyst for writing this down.)