It's tough to gauge really as Marner struggled early on but how much of that was a combination of playing with Ritchie (who didnt even work when he had both Matthews and Marner together with him) and an injury issue? Even looking since 2019, there just isnt a whole lot of data for Marner away from the top 2 centers and even when there is, he's being used in a defensive shutdown role which is gonna skew things quite a bit. Marner's Ozone starts away from those guys are in the mid 30s in limited minutes (but he still has a decent 50+% GF% which is kinda nice)
Just looking at the 5v5 data for Matthews with and without Marner during the Keefe era:
Marner/Matthews:
TOI: 1421:13
CF%: 55.18
xGF%: 61.34
GF%: 63.80
Matthews without Marner:
TOI: 687:03
CF%: 54.55
xGF%: 55.05
GF%: 53.79
With Ozone faceoffs/usage being almost identical and a decent sample size, this is a pretty good indication that they should be kept together. Looking at a comparable league-wide to Matthews/Marner/(fill in the blank) with the same type numbers and you can probably count it on one hand. As a line, they have pretty much the same numbers as Boston's "perfection" line.
Looking at Tavares/Nylander over the same stretch and we get:
Tavares/Nylander:
TOI: 1163:19
CF%: 54.71
xGF%: 58.12
GF%: 54.51
It's hard to compare Tavares in the same way as we did Matthews as he also played much more defensive minutes away from Nylander since Keefe came on board but Nylander did play similar usage away from Tavares:
Nylander without Tavares:
TOI: 597:28
CF%: 54.01
xGF%: 54.61
GF%: 47.75
Pretty big difference, especially in the outcomes. Pretty sure we are going to be seeing more of what we have been seeing the last couple of years because it just flat out works.