We don't have the actual data but whatever the previous results were we can normalize them so the average score is 50%. We also don't know how many students there are, let's call it 100, so 97 pack students and 3 lone wolves.
We have 20% better results for the pack which received a uniform score. Let's say that's the 70th percentile. (This is a simplification we'll have to make do with until the author publishes the actual score data.)
One lone wolf scored at that 70th point, one below, one above.
The 70th percentile represents the best work that the pack of 97 could agree on. Out of the lone wolves, 2/3 of them did as well as this or better.
It is not hard to imagine that there was better potential work than 70th percentile among at least some of the 97 individuals in the pack than what was chosen through committee, work that will never be manifest because a committee is always a compromise.
It is also not hard to imagine how it is that the average went up given that the half of student population who previously scored in the 0 to 50th percentile are suddenly cast into the 70th percentile simply by choosing to join the pack, even though their contribution, insight and understanding remains exactly the same as before.
It is not hard to imagine a scenario where the failing students who have learned nothing and acquired no advancement in their understanding and yet who just scored 70% instead of 0-50% still have no insight. It is not hard to imagine that the really excellent possible work was suppressed by the groupthink of the crowd. Yet the "average" score is now 20% higher.
We don't even have to consider that this single question essay exam was almost certainly graded in a completely subjective manner and thus the percentages don't mean anything objective at all.