Ben, interesting and nice work which also shows the variation in seeing from night to night.
However, I would say that 2.5s, 3.0s and 3.5s are within the margins of error of each other. More data is needed to be conclusive as the night-to-night variation in seeing is very significant, just look at the spread of points at 3 seconds. As an example, from the graph one could even try to argue that 2.5 seconds is better than 3 seconds.
Also, probably the final results will also depend on the location, some places will generally have better seeing than the others.
Having said that, the trend for shorter exposure times is very clearly demonstrated