The primary issue here is you're using "PP%" for different units. What does that even mean? If PP2 comes out for an 8 second shift are you going to say they are 0/1? In that case the average success rate for a PP2 is going to be very low, likely below 10%. You very obviously need to look at things on a per minute basis here. Even that isn't perfect because the success rate of PP2 getting a rare shift to start a PP will be far higher than their success rate when getting the second shift on a PP.
Second, the Rangers didn't have any forwards (other than Bonino who was cut) average over 2 minutes/game on the PK so you're talking about such a minimum about of ice time that you're cutting out per game. The Rangers were shorthanded 2.83 times/game. Zibanejad played 1:54/game on the PK. So he averages about 40 seconds per PK. Cutting out one PK, so 40 seconds of ice time is going to make some massive difference? He's also an extremely good scorer on the PK so cutting his ice time there is stupid. It's not entirely unheard of for players to score at a higher rate at 4v5 than 5v5.