the hand calculations are based on the part staying the same shape. All of your other parts are symmetric so applying the load the center is still symmetric, but with an L-angle the lack of symmetry places the neutral axis away from the center of the face. The simulation solves for the fact that your L will start to roll because it is stiffer on one side than the other. If you mirror the L into a U-channel it should be correct because of symmetry places the force coincident to the neutral plane again.
Yes I have mirrored a cross section that was angle shaped & bending out of plane into a channel cross section. I used the same part & simulation study with the same loading & restraints & after mirroring it to a channel it bent as expected--all in plane & matching hand calcs. I tried this with a channel in both directions--as a "C" and a "U" relative to the load in the Y direction.
I figured this was probably the case but could not find anything in my text books to warn of this in hand calcs.
The simulation results for the angle show more bending stress stress than the hand calcs along direction 2. The simulation results also have a good amount of bending stress in direction 1, where as hand calcs obviously have zero with the simple Stress = M/S equation.