Notwithstanding the fact that kr data from one sample does not a model make I'll leave you with the following:
I'm no kr expert, but I have built/matched many models over the years and intuitively I'd say don't do it. My rationale comes from considering a waterflood for the fractured sample. As it's flooded, the water will likely zip through the fracures leaving behind a lot of oil in the matrix. After water break through, the average Sw may not have changed much in comparison to the non-fractured formation, but obviously the water cut/krw has and won't follow the same krw/Sw curve.