My first thought is that it will work, but the height might be only accurate to 1km, or maybe 10km. The change of the 3D magnetic vector does not change that much, and the sensor is not that accurate. It is just my feeling telling me this, so I might be wrong.
ADDED: I might have misunderstood your question. You do not want to know the absolute height ? But only the orientation of the heliostat ? The azimut and tilting ? That is not only possible, it will work very well I think. You don't have information about the rotation around the magnetic vector as jremington wrote. But the heliostat only rotates around 2 axis, the 'unknown' axis is not used, as long as the top of the heliostat never tilts sideways. So I think that all the information is available to calculate the orientation of the heliostat.
A magnetometer is sensitive for magnetic disturbance by wires and metal, and a accelerometer is not.