Yours is an interesting proposal. To show how interesting it can be, let's assume that an ion rocket will be attached to 2004 MN4 tomorrow.
Where will we attach it? The asteroid is undoubtedly rotating, creating a monumental problem in determining where the rocket should be attached in order to force the asteroid away from Earth.
Over the next 25 years the asteroid will travel about 14.5 billion miles. From that distance (imagine the asteroid's orbit is stretched out into a straight line 14.5 billion miles long) target Earth subtrahends an angle of about 0.1 arc seconds. Out of the uncountable possible scenarios, one is if the persons deciding where to place the ion rocket make an error of 0.1 arc seconds in directing the force of the rocket in the proper direction, they may cause the asteroid to hit Earth rather than miss it.
It might be less risky to attempt to blow the sucker up into lots of smaller, less threatening chunks, IMO.
:') An alternative is to use the ion drive to speed up the rock at perihelion to alter its trajectory. Any rotation would be addressed first.