Hello,
you can set up fitness function for galapagos like this:
It will multiply fitness number by 100 if deflection is overreached and therefore disadvantage the gene in galapagos so it should be kicked out of pool.
For the other things… Pay attantion to buckling lengths! As default, they are considered for lengths between connection points, not between your lateral supports. You can tell by checking details on your calculations. Ncry = Ncrz which means if symetrical cross section is used that buckling lengths are same in both directions. Correct me if Im wrong… @karamba3d could comment on that.
As from the previous post, it seems you could model lateral beams in the position of lateral supports to get the correct buckling lengths. But I would recommed to verify by checking on details of utilisation for buckling lengths.
Other option would be to use buckling mode as initial shape for structure , calculate with second order effects and check on stresses.
This might be affecting your optimised geometry.