This appears to be conventional wisdom, but one need only look to the graph you attached on the previous page to realize that this is false. A small difference in the offensive versus defensive zone starts is not going to affect corsi in any significant way. The effect of which zone a player starts in following a faceoff is severely overstated, and has become an empty talking point on these boards.
Lets look at it like this. Why do you think that players with the highest defensive zone starts struggle to have their CF% over 50%? Why do you think that players with the highest offensive zone starts have CF% consistently well above 50%.
Then if we agree that worse players can be better possession players in better situations, then what are we actually trying to argue. Is the hypothesis that a collection of the really good possession players make the best team? Take a bunch of guys that are in theory undervalued, but have great CF%, and create a team of those players? If you don't believe that then possession stats at the core are kind of meaningless.
The counter argument to that is that the top Corsi teams typically win. Well yes, that is obviously true, more shots likely means more chances and likely means more goals. For a team, Corsi can tell you who is good and who is bad. For an individual, it does not give you that same conclusion. If you take a collection of good possession players, it does not mean those possession numbers will remain the same or even remain good in this hypothetical possession team. I'm not talking about a collection of Doughty's, I'm talking about a collection of the Corsi greats that don't pass the eye test or have favor variables for their high corsi, guys like Jackman from last season or Ribeiro or Brown.
That's the core of the argument here, if the Corsi works for teams, but not for players, then it's not a good predictive measure. It only works when you can truly balance out the variables, when you combine the favorable variables and the unfavorable variables of individuals and combine them. Corsi fails as a predicative or evaluative measure for individuals because it assumes that Pietrangelo and Shattenkirk would have the same Corsi in reversed roles.
That's why people on my side give Corsi and Hero charts very little credit. This is really the only area in this conversation where we can have a healthy debate. Everything will just be disagreed on and there will be no meaningful discussions coming from them.