How to simulate an impulse force acting on my system? What step-size and solvers are best?

Let me describe the system first: I have a cube satellite of size 10cm, which has four very small thrusters located at four edges of one of the surface (all pointing in one direction). Firing of these thrusters can only produce torques in two dimensions, so my task is to control two angles (lets say yaw and pitch). I have a dynamic simulation which fires the thrusters according to simple PD-like control law and this works fine in my current setup. But when I change the step size, the settling time changes by the same order which is something not expected. The thruster (actuator) block is connected to plant (SAT Original) block which simulates the change in angular speed based on the last firing like this.

The actuator block is running at 100ms, such that i can have two different torques inputs at 10.0 and 10.1 second. In real situation the reaction force executed by the thruster is an impulse which lasts only tens to hundred of microseconds. I am running the simulation at 25ms or 50ms or 100ms. But seems like changing the stepsize of the solver can results the same force being executed in the simulation for the same time as the stepsize of the solver which seems rather wrong. I am using fixed step ode3 solver. I want to understand a few things

1. If i run the simulation at 100ms. Does the SAT original block sees the force/moment as a continous signal which is rather a train of impulse at every 100ms.
2. Otherwise how should i correctly generate this impulse force acting on the satellite?

The inside of "SAT Original" looks like this (mainly integrators)

As i mentioned before, i get different settling times when i change the step size which i think is wrong.

Thanks for the responses.

Regards,

Sid