I wonder if it is a "wide spec" material, where the MFI could have variation and still be "in spec". Can you correlate the defect to specific melt flows?
I had a similar problem where material had to be at low end of spec to produce acceptable parts.
Just a thought.