I look at the FULL data from the link I provided and cannot draw any such conclusion. At first the M1 filter did look to be a bit better using Blackstone's particle count method. But by the last run, it's clear that the test is either flawed, the measurement method funky, or there is no statistical difference.