Not sure about this, but one possiblity could be the fact that the solids are restrained on faces, and the beams are restrained on joints. If the restrained faces are sufficiently large, they will react some of the bending moment caused by the load. Beam joints, on the other hand, are points that are free to rotate. This may be allowing them to pivot more freely, giving you more deflection. Try shrinking the size of the fixtured split faces in the solid model; the deflection might increase.
Thanks for your respons.
My VAR solved this issue. The fault is mine and happened during the setup of the beam version..
The "on flat surface" support on the beam version is not enough (beam element has 6 dof). I also need to prevent rotation in the axis between the joints to simulate the bending moment that occurs in the solid and surface solutions. When I add this restraint the solution gives similar results also in the beam version.