I try to see the video, but is not running, please explain your problem with moredetails and we will try to help you, thanks!.
Please upload a copy of the FEMAP model, this way I will see details like mesh size used, element orientation, etc..
There's a big difference in these two models. The vertical load on the curved beam introduces a torque loading on the beam. In the straight model, you do not get any of this. Here's a plot of the torque values in the two models -
I worked on the V-22 Osprey Prototypes, and the wing was attached to a circular ring on top of the fuselage. There were four connections between the wing and this ring. As the wing was rotated, the sliding shoes got away from the supports and torqued the cross section so much that most of the deflection was due to the torque.
It takes to my attention as well the differences in results between both models. First at all is important to say that both models are different, and so the results are consequence of the changes imposed to the geometry.
For instance, due to the use of straight edges the model#2 do not experience any torque internal forces, where Model#1 has some values along the beam.
Also, if we plot bending results along both beam models we can see that Model#2 with Straight Edges exhibit bigger result values of Bending Moment, almost double in the middle between supports, from around 22 Nmm to 39 Nmm.
I don't have an explanation of why in Model#2 the bending moments are zero in the supports, and reaching maximum negative values of 37.89 Nmm in Model#1. I ran the model as linear & nonlinear and results are similar, then we can say that different results are consequence of difference geometry.
Regarding beam stress results, Model#2 with straight edges experience bigger resultant combined stress results than Model#1 using perfect arc geometry, the lesson to learn is that a perfect arc geometry is better than approximating the arc with straight lines, bending moments are smaller and then the final results as well.
Regarding REACTIONS & DISPLACEMENTS we can see based in the results comparison that Model#2 exhibit bigger displacements. Both models has applied the same vertical loading of 50 N each in total.
Again, I think this all due to the introduction of a torque due to the curved beam hanging out past the axis of the point to point constraints. With only translation constraints, this torque is reacted as forces at the constraints. This results is an uneven distribution of vertical forces along the cross section.
In the straight model, the end results is five pinned-pinned beam models, with zero bending moments at each constraint.
In a model with an uneven distribution of reaction forces, (see Blas' Free Body Diagram), bending moments have to be transferred across the constraint, even in a simple straight beam model, with unsymmetric/uneven loading, you will see this -