Negative Permeability

I am a graduate student working on a project dealing with core formation in planetesimals. I am using Palabos and the LB code to determine permeability of FeS melt through a silicate matrix (from binarized X-ray microtomography images) through increasing steps of deformation. There is only about 4% FeS melt, which is not quite at the percolation threshold, so we expect the permeability to be low, but increase with increasing deformation. But some of the permeabilities end up being negative. We were thinking that this was because the permeability is so close zero. The highest permeability we obtained was in the e-04 range.

Has anyone else encountered this before or maybe know the cause of negative permeability? At this point, I do not know much about the LB code, but I can provide as much additional information as I can to maybe get a better idea of what is going on. Thanks!