In case of MD simulation starting from a crystal structure we used to deviate a lot from the initial starting conformation. And the traditional way to calculate MM/PBSA or GBSA based free energy calculations is used to get snapshots average over certain time-scale. It is possible that these snapshots include poses which are much deviated from the original starting structure (crystal structure). Will it be better to take in account snapshots closer to the crystal structure say for example within ~1.5A deviation from the starting structure and to take all those snapshots to calculate MM/PBSA calculation.

Is it a good strategy?