Adding a capacitor anywhere (either series or parallel) over speaker leads is only going to handicap one amplifier with respect to the other. It'll either act as an HPF (series- high value) or a LPF (parallel- low value).
EQ'ing out frequency response anomalies makes the entire "test" useless. So, he made the amplifiers sound the same, excludes the amplifiers with residual noise (all of them), handicaps the powerful ones and wonders why nobody can claim the prize in his stacked 2x12 correct ABX challenge?