I've not the read full thing, will have a poke around when I can.
I think you hit on the crux of it at the end though Bright Star, that all any exam can be is a surrogate marker. Now, MRCP results against medschool attended is interesting in of itself and the data mentioned in the abstract is worth following up on. Following patterns, particularly in the London schools as they close, merge and shift their curricula from fairly uniform to about as diverse as they could get would undoubtedly be interesting. Seeing as this is an exam a significant number of graduates have to take and the only way into senior medicine, I would've thought it'd be something measured as a matter of routine rather than exception to be honest.
The danger is trying to use the data to say anything more than that really. Defining being a 'good doctor', even if you limit it strictly to clinical competence and completely ignore colleague and patient communication and teaching skills, is a ridiculously slippery thing to try and do. Its that last bit that gives me a kneejerk aversion to putting all the weight behind a national license. That being said, there isn't a medschool model I've heard of that doesn't involve a Great Big Exam of some nature and whenever you do that you get people teaching and learning to it, so maybe it won't make all that much difference.