So what about the public analytics that have his two-way play as rather putrid?
There are two sets of metrics that people are looking at: On-ice and individual. They paint very different pictures. You are referring to on-ice metrics, but most people are referring to his individual metrics.
The biggest reason for Celebrini's rough
on-ice two-way metrics is that Celebrini was out when Jake Walman - SJ's true MVP this season - was in the lineup, and then Celebrini returned right around the time that Walman got hurt. If you look at the
WOWY, it still does not look great for Celebrini, but it paints a much more neutral picture - the team with and without Celebrini on-ice is basically the same once you adjust for the presence of Walman.
This is not good, considering SJ is quite a bad team, but I think putrid is clearly an overstatement. Now, the biggest reason that his on-ice metrics are still not "good" even after accounting for the context is, IMO, puck and shift length management. Celebrini frequently makes hope plays that feed the opposition rush; by no coincidence he leads the team in giveaway rate. He also frequently takes extended shifts and gets hemmed in at the end of them. These habits are already improving and should be ironed out by season's end.
On the flip side, Celebrini's elite
individual two-way metrics speak clearly to his abilities, and are much more projectable long-term than on-ice Corsi or xGF over a 14 game sample. There is a very good chance that Celebrini will have strong on-ice Corsi and xGF in his prime; what are the chances that Michkov or Bedard will ever lead the NHL in puck battles won per game, even over a small sample? Quite low. And players who dominate in puck battles tend to be strong two-way players; 5 of the other 9 players on
this list have been finalists for the Selke trophy, and Bergeron was usually near the top of the league in puck battles when he was still playing.
It is perfectly reasonable to cite Celebrini's projectable defensive skills as a reason to take him over Bedard and Michkov. (Especially considering the other two also have less than stellar on-ice two-way metrics.)