Using a design study to determine failure given varying geometry (nonlinear)

Hello,

I'm currently doing some sensitivity analysis and am trying to automate analysis runs so I don't need to create a seperate file and study for each model dimension change.  Essentially, I want to determine when a column will fail given certain section loss at the midpoint.  The failure mode is a mix between Euler and local buckling. 

I've created a design study trial with varying dimensions that I want to investigate, but the results I am getting are not consistent.  The scenarios all mimic the original nonlinear analysis and use the exact same number of steps.  Given the geometric changes in "Scenario 1", the end time step is 1.644 seconds indicating a failure load of 644kN, however when I do the same run using an independent analysis I get an end time step of 1.602s indicating a failure load of 602kN.

Attached are the files.  I'm using the nonlinear analysis in check.zip to try and validate Scenario 1 in design study.zip

Cheers,

Andrew

SolidworksSimulation