Yes, unfortunately the postprocessing commands aren't written to the ds.dat file. Only the preprocessing and solution commands. I'm quite sure Mechanical is doing something like spoint/fsum. The centroid is probably computed based on the average node location. You could double check it with a commands object using APDL commands, but I doubt it would differ very much. You're assuming the stress varies linearly through the top flange thickness and the section. Is it possible that assumption is introducing a discrepancy?
You are navigating away from the AIS Discovery experience