I have a set of low resonant protein complex models generated from global docking. I did local docking on each model. Some of them have given the "energy funnel" with its lowest energy structures located around a point, but some don't as the low energy structures are dispersed over the plot, which I guess is a sign of failure in docking. Would this suggest that my starting structures are not good, I mean, not energetically favorable? All these starting models are predicted from XL-mass spec results and a global docking program.
It's somewhat expected that you won't get 100% good results with global docking starting structures. Typically global docking uses a lower resolution scoring function and a coarser sampling method to allow for quicker scanning of docking possibilities. It's not as accurate, but it's much faster. So normally you'll get possible starting points with global docking, and then try to refine them with local/high-res docking. Some structures from the global dock will be false positives, which will "fail" by local docking, but hopefully some of the structures from the global docking can be refined by the local docking to a structure that is reasonable.