Calculating the load distribution plot is a post-processing step -- the solutions (between the two file formats) should be identical (unless the wake was detected differently.
The experimental file format contains not only x,y,z information about each point, but also U,V information about where the point is located on the surface of a component.
The spanwise load integration algorithm uses that U,V information to do the spanwise load stuff. If it does not have it, it does the best it can with just X,Y,Z.
All that said, load distributions on thick-surface models are a very new feature that was added in the most recent version. It looks like there are still some improvements to be made - particularly in the version that relies on only X,Y,Z.
Specifically -- I don't know why it influenced your results (with the center line only change), but I'm also not super surprised.
Rob