Eh, it's not that unreasonable. Comes down to consistency and scoring timely goals when the game is on the line.
For instance, (I'll bump that up to a 30% PP for ease of math) the 30% PP unit could do a lot of its damage in one blowout win, going 3 for 5 in a 6-1 win, but then go 0 for 5 in the second game in a 2-1 loss. At the end of two games, it's operating at a 30% efficiency.
Meanwhile, the 20% PP could consistently score 1 goal in each of the two games, going 1 for 5 in both games, contributing to a 3-2 win and a 4-3 win, where that 1 goal was the difference in each score. Yet at the end of the two games, it's operating at a 20% efficiency.
Small sample size, but that's an example of how the 20% PP unit was more reliable than the 27% (in my example the 30% unit) PP because it showed up when the team badly needed the PP to be the difference in the score.