IIR statistics correctly, say two players have the same/similar average ratings over the same # of rated rounds. One player has ratings that run the gamut from 820 - 950, while the other has ratings that are all pretty closely packed around 890.
The player whose round ratings are more consistent would have a smaller std deviation than the player whose scores are all over the place. Hence, the threshold for one of their rounds being dropped is tighter than the threshold would be for the less consistent player. So, the more consistent your rated rounds are, the easier it is to drop a bad round.
Essentially, rounds are more easily identifiable as statistically aberrant when the data is closely packed together. Spreading round ratings over a greater range means there's greater deviation, so it's tougher to say a low score is truly a statistical outlier.
I hope that helps for those unfamiliar with how std deviation works.