I've noticed that xG isn't as instructive in these samples because they're so small. The stat tends to overrate shooting% luck in < 200 minutes, where combos with good GF% have poor xG and vice versa. For example, those 3 combos you listed had just 39% of goals scored. Kane-O'Reilly-Gionta was weirdly good at everything though (50% GF, positive corsi), so maybe there's some untapped juice in a say, Kane-O'Reilly-Griffith trio.
Gionta had really good positioning, the mobility to get to his spots in the offensive zone, and was fine with being a glue piece, so those are probably the ideal traits to look for in his replacement (again, Griffith scoring highly here).
Also, for the moment, it seems Housley is putting consistently poor exhibition play from the Kane/Pominville tandem on Poms. Kane/Eichel got going once Griffith spotted up a line, so it'll be interesting to see how long he holds that belief.