results depend on parameter sweep order?

Hi all,
Is it possible that in some cases the results of a structural mechanics simulation depends on the order in which a parametric sweep is done?

I am simulating the deformation of an object when it "hangs" from its support-points. The support can be tilted wrt. vertical about +/- 2 degrees.
So I am running a simulation where the body-force depends on the tilt-angle = range(-2,0.1,+2)
From this I would expect to get symmetric results, i.e. the -2 degrees tilt situation should look exactly symmetrical compared to the +2 degrees tilt situation.
However I seem to get "skewed" results and the -2 and +2 degrees tilt results do not look symmetrical.

What I think is happening is that comsol solves the FEA problem for one case, the first parameter value of -2 degrees, and then all other solutions are small (linear?) modifications to this original solution.
I have yet to try what happens if I run the same simulation with tilt-angle = range(+2,-0.1,-2), or with individual parameter values of -2, 0, +2 degrees of tilt - allways erasing the previous solution between runs.

My model consists of a piece of glass, and the deformations are small, 1e-12 to 1e-10 displacements, so there might be some numerics that should be tuned also..

thanks,
Anders