Yes, when you model half a unit cell, the force output is half. You can simply double the value to get the expected full unit cell force.
I hope there will be fewer problems running the transient analysis on the 3D shell model compared with the 2D plane stress model.
You could compare F vs D for the 2D Plane Stress and the 3D shell model to gain some confidence that they are similar.
Isn't showing the F vs D graph a way to show progress in the internship?
A big benefit of the 3D shell model is that it will solve much faster than the 2D plane stress model because there are a lot less equations in the matrix, so that will help in checking that the Transient model is correct. Another change that might help the model to converge is to switch to a linear elastic material from the hyperelastic material. You can see if this makes a difference.
Regarding the response of the 2 Hz model using 2D plane stress, the direction is correct. Looking at the model results more carefully, the displacement of the top mass due to gravity did not compress the springs to the neutral point, so when the base motion started, it was not in the null position. That is why the top mass had a large response.
In a SDOF dynamics model of a mass on a spring, it can happen that the motion of the mass can exceed the motion of the base. This is expected when the base is oscillating at the natural frequency of the system.