I am performing Steered molecular dynamics simulations and trying to get PMF using work distribution curves. The problem is when i am tried to look into literature for such calculations, everywhere i found work distribution for backward sampling (say refolding of a protein) lags behind (average work value is lesser than forward average value) the forward sampling (say unfolding of protein). On the other hand in my system which is simple NaCl in water (test system) i found at certain distances forward work distribution is ahead of backward and vice versa. Can i get some help ?