In a nonlinear static two-step subcase, why does the load step decrease after the first step?

I am using MSC Nastran for nonlinear static analysis (sol 400).

Step 1 contains a small perturbation load laterally on a thin-walled cylinder. Step 2 contains a very large compressive load.

Usually when I do this, the Load Step increments from 0 to 1 for the first step. Then for the second step, the load increments from 1 to a number under 2 (basically until it stops because the structure fails.) So in this second step, if the indicated load step is 1.3, then the actual compressive load in that load step is 0.3*applied load. (0.3 because take away 1.0 as that is from the previous step.)

The problem is now I am in a situation where the load step starts decreasing the moment the second step starts (right after LOAD STEP = 1.0).

I am using the arc-length method, so the load can decrease, but it shouldn't decrease this early. I want at least some compressive load until it starts decreasing. Because my understanding is that the compressive load in the second step is negative, and is therefore actually causing tension instead of compression.

So does anyone know what might be wrong and how to solve this problem? I would appreciate any kind of help, thank you!