Optimize structure

I met some problems in structural optimization. I used the MP recommend input sets, but get different results(e.g. total energy). And, I try to re-submit the optimize structure task based on previous calculation, I got the right results! So, I guess that twice calculations are need for optimizing structure. Is that right?
The figure is a screen-shot of MP web.

Yes, our optimization procedure uses two optimization steps to minimize the effect of Pulay stresses. More specifically, it uses custodian’s “double_relaxation_run” feature.

Got it! Thanks for your reply.