I have done quite a bit of work (both as an external consultant and as an internal consultant) in the area of developing behavioral competency frameworks/competency based development. I still feel that they are useful in particular organization contexts. However my current thinking on competency models is to consider them more as an 'intermediate stage' in the evolution (of capability building efforts) rather than as the final stage.
One of the basic assumptions behind developing a behavioral competency model is that there is one particular behavioral pattern that would lead to superior results in a particular job(i.e there is 'one best way' to do the job). This might not be a valid assumption, in the case of most of the non-routine jobs(where it is not desirable to specify, in micro detail, the exact procedure to perform the job). Thus in the case of non-routine jobs, competency models might lead to 'theoretical'/'copy book style' behavioral prescriptions that might not be optimal for effectiveness on the job.
However, competency models (like copy book style training) has its own usefulness in particular contexts. It is useful to teach the beginners the 'copy book style' of doing a particular task, with the understanding that once they master the basics they are expected to improvise and modify the 'copy book style procedure' to suit their personal style and the specific context/ particular task at hand. Let is look at a simplified example (outside the work context) - cricket coaching. As part of coaching kids/beginners they are taught the copy book style of playing each shot (e.g. how to play a cover drive, sweep shot etc.). However, if we look at established cricket players, most of them play a (slight) variation(s) of the shots based on their personal style and the demands of the situation.
Thus I feel that we should use competency models and then go beyond them !!!