That's it, I think I understand why for a given surface, by replacing the equivalent force calculated by FreeCad with a displacement (null here), the results do not match.
In fact it's logical: FreeCad calculates the total force by summing the forces calculated at each node. And when we apply this force (total result), FreeCad applies this force by distributing it only on a few points that we do not know (probably a few key nodes or only one node, in any case probably not the same nodes as initially and probably not with the same distribution of forces in the starting nodes).
So the result of the deformation of the piece is different at the arrival. There is also an influence on the results of the other displacement constraints.
There is also the influence of the fineness of the mesh, but this is another problem.
Today we can apply a force on a surface or edge, therefore without knowing the force equivalent to a given displacement. This generates a certain amount of deformation.
If we apply this displacement, it generates a different deformation and vice versa.
There is no effective correspondence between these two input conditions, and above all they give different deformations when they are supposed to give identical results.
Wouldn't it be useful to create a tool that allows to apply the true force equivalent to a displacement calculated by FreeCad? Or at least in the form of a list of forces to be applied to nodes in script for example via FEMMeshGmsh.inp
Example of an aberration of forces equivalent to one zone:
Sum of FreeCad equivalent force : 3,326878 N