Reviewers say that there are so many variables in trying to compare stabilization systems that it's very hard to test objectively, no matter how assiduous you are about them: camera shake can involve a whole lot of factors, not all of them even related to the machinery itself: the only way is to make a whole lot of images at different shutter speeds and score them subjectively, then show the probability of sharpness at each: that's not even taking into account the various ways that camera shake can manifest: is it a sudden jolt, a slow sway, a slow or fast vibration, where in the camera-lens combination is there the most motion, ...then things like how well does the system respond when there's no or *less* motion. Does it overcorrect when the shooter's very steady on their own?
One system might be better at any number of these combinations than the next, so even if you could test all the cameras exactly the same way, it might not indicate what a given person might get in practice.
You're probably better off asking a whole bunch of people who've tried both types. The general consensus seems to be there isn't a huge difference in practice. I've only used the in-camera kind, myself, but my understanding of the consensus is that the in-lens type may be a bit more powerful to some folks' estimations, but the difference in practice isn't huge. For someone like me, who's pretty steady on a good day, anyway, and mostly needs the SR for lower-light stuff with shorter lenses, I think it's the best solution, especially since no amount of money, even if I had it, can buy my favorite FL's I use in primes.