Do you have an experimental stress-strain curve for the steel that the rod is made from? Maybe the bilinear isotropic hardening material model is not a good match to the actual material data.
What flaws were in the actual sample that were not in the model?
What were the measured dimensions of the actual sample and how much do those numbers differ from the model?
Where did the actual sample fail? Where does the model predict the failure occurs? If the actual sample failed at the threads at the end, then you can adjust the model to include them.