What I've seen for problems in Root Cause Analysis basically boil down to what are you doing a Root Cause Analysis on? Here in the Department of Energy, I've seen most folks focus upon individual events, individual symptoms and do a "root cause analysis" on them. Because they are not looking at the system that created the individual event or symptom, the RCA is doomed to failure because it is looking at too micro of a level.
Another issue with RCA's that I've seen is that people doing them don't segragate the issue they are looking at into stable and non-stable systems. So often a "trend" is declared on 2 or 3 events and an RCA is done, without recognizing that the so called trend is the result of a stable system, and again, that implied the whole system should be looked at. Now, if there is a statistically significant trend on a control chart, then an RCA of that trend should provide benefit.
I suppose it all comes down to most RCA's are a dissection of an issue, rather than an integration as part of a larger system. To illustrate this (and borrowing from Russ Ackoff), this is like doing a root cause analysis of hand pain by chopping off your hand, laying it on a table, and trying to determine the source of the pain. As an aside - how many people dissected a frog in science class, and then learned how to put it back together again?