At the peak of the ordinals spikes I ran several feerate comparisons on filtered and unfiltered.
I detected no statistically significant variation in feerates. All of them either were sufficient to get into the new block, or both standard and filtered estimates were wrong due to a sudden spike in demand.
