Great!  I hope you see how much easier it is to interpret the results by plotting deformation instead of velocity.

Configurations 1, 2, and 4 give garbage results.

Configuration 3 has some potential to give good results but the initial motion seems corrupted. Make sure the contact is Near Open before the simulation starts. There should be a clean parabolic curve of constant acceleration for a part falling from rest like I show below. You can see a parabola for Step 3 of your simulation.

Double the end time of Step 1 so that the part comes to rest before the rapid acceleration reversal step, then make sure the end time for Step 3 is also doubled so that you have time to see the bouncing and the part come to rest.