The problem is that StructTemperatureLoad (and Eigenstrain loads) are not actually loads, they have no direct contributions to the external forces vector.
Temperature load will modify the response of the material model (which will subtract the thermal expansion from the strain in order to obtain only the elastic strain), which in turn modifies the internal forces vector, as opposed to the external one.
Even if there is just a simple linear thermal expansion, it effects the rest of the (possibly nonlinear) material response as well.

In the case of a linear material model (or at least, a model such as isotropic yield surfaces which do not depend on a large volumetric strain), it would be possible to put the thermal expansion in the external forces vector (thus enabling for indirect control).

If the material subject to thermal expansion is just some linear elastic model, I would suggest this approach, if not, then I'm afraid I don't have any solution.
A new type of load would need to be added; something which applies "alpha * dt * E" on as a volumetric load. Support for different element types must be added manually as well. However, it must be made clear that it should never be used together with nonlinear material models, since it adds a large artificial "elastic" strain.