Frame structure optimisation with Millipede

Hi all, I'm trying to optimise a frame structure using Millipede. The overall goal is to adjust the thickness of each member so that it corresponds to the stress which occurs there. 

First I create a basis FE model to get some stresses to evaluate and then I remap these stresses into a thickness within a certain domain. These values are then used as input to a new FE model.

I look at a cantilever beam (8 x 3 m) which has uniformly distributed loads on top and bottom and all DOF's are fixed in the end (left side).

I get the exact opposite result as I would expect. See the result below.

I've tried to switch the domain input of the thickness which kind of gives a result more likely to what I would expect but still does some weird thing at the end of the cantilever beam. Also I don't really see the logic in doing that.

I've also tried the region boundary support instead but with no luck.

Can anyone see what I'm doing wrong?

Thank you very much!