Increasing n gives increasing weight to outliers, to the point where L-infinity is just the single maximum value. I would guess this is less useful when attempting to understand how existing data can make conclusions about future aggregate/typical cases vs. analyzing specifically the outliers.