Hi Jeremy,
Looks to me like an initialization issue. I don't have R2020b on my computer so I tried this on R2021b, the oldest version I have.
Change that PS constant to a step + some filtering. You can do this as PS step + PS transfer function, or as Step + Simulink-PS Convert (with input filtering). I did the latter and it seems to give reasonable results.
Some other suggestions, but totally optional.
(1) I personally found that gas-charged accumulator a bit confusing to set up, especially the initial conditions. I prefer the Tank (G-TL)
(2) since you are using a torque source, it may make sense to attach an inertia to the shaft, just to avoid any chance of F=m*a while m=0.
