The best case performance of the thing would require that the engine thrust equals the amount of force to be exerted on the asteroid, but unless the distance between the craft and the asteroid is large or the craft uses a pair of engines firing at extremely shallow angles relative to the asteroid (which would be extremely inefficient) I would think the ejecta from the engines would push on the asteroid in such fashion as to work against the desired motion.
Agreed, but how about this?:
Replace the spacecraft with an equivalent inert mass. Tether the mass to a platform via a very long (~1000 kilometer) cable. Mount four, or better yet, six engines symmetrically on the platform so as to exert the necessary force keeping the inert mass at a constant distance from the asteroid. The angling of the engines can now easily be made shallow enough that their ejecta will miss the asteroid while providing a reasonably efficient thrust.
All of this technology is doable now (at admittedly enormous expense). In at most a few decades, it will be cheap enough for serious consideration.
This would seem like an improvement on the concept, especially since the "inert mass" need not be of terrestrial origin. In addition to improving rocket efficiencies, this approach would also allow the mass to be much closer to the asteroid (indeed, I should think that having it land on the surface would probably not pose any particular problem).
I'm still unconvinced, though, of the practicality of getting any useful amount of mass into the right place to do any good. Gravity is a pretty weak force, after all, so the mass in question would have to be pretty incredibly big to do anything.