The last column has a stat error > 3%. If there is an effect vs the
integrated purity, it seems to be due to the usual bias in yields estimate
with low purities. The stat error associated to the calculation of this
number should cover this systematics.