Comment by aakilfernandes

Could this just be a result of low sampling by the author? If you reduce a sample to only include some tiny edge case, the resulting data points are going to be weird in random ways.