Ok but deductive reasoning would tell you that those players who have the most defensive zones starts are going to have the weakest Corsi, playing against the strongest competition generally. And having the primary responsibility of not only clearing the zone but advancing the puck into a position where they can shoot. That would take at least what 25 seconds out of a 50 second shift? Easier to have a higher shot rate when you are in the offensive zone and win the draw?
Deductive reasoning would tell you that those things would impact goals too...
If it's easier to out shoot, it's also easier to outscore...
So that hurts your numbers, just like mine.
Mine don't have the problem of:
* small samples
* goaltending impacts
* not being TOI sensitive
* not having stupid situations like PK +s and PP-s and empty nets thrown in randomly
HOWEVER, you are also highly overrating those factors.
1) Zone starts impact is actually 7 seconds on average, and also well over 50% of shifts start on the fly. Also, we've discovered that bad players tend to cause a lot of their D-zone starts, like Stuart, who have high numbers of failed chip-out entries that turn into icings or being trapped in their own zone and forced to ice it for a breather.
2) Almost all players from top to bottom see opponents equally amount in terms of percentage of their icetime. Both coaches are trying to get their line matches and also on the fly changes happen all the time. As a random example: a "sheltered" player at the end of the year may see 1st lines 31% of the time, while a "tough mins" player at the end will see maybe 33% of the time.
Also, we've known about these factors for about 10 years, and have worked towards diminishing their value anyways... although it's been shown to need very little adjusting outside of linemate factors.