ddG calculations to study point mutations

Hi,

I am using Rosetta protocol to determine ddG changes on point mutations in a peptide bound to a protein complex. I first used FastRelax to remove any steric clashes in the protein structure and then ran the ddG protocol to generate ~1000 structures till there is no change in ddG values. I now want to build a correlation plot between the Rosetta ddG values and experimental calculations.

My question is what ddG value should I consider for my calculations. I am presently using the min ddG value presently. Should I consider the value at which it is converging or the average ddG of ~1000 structures?

Thank you.

Nikita