I'm happy for Easy Allies if they believe they can use the numeric system better. I liked the stars system but agree with their arguments for the change. I think the traditional numerical 1-10 scale (or rather, 6-10) is not the best solution, though. If there is a problem with reviews now (the fact that scores mean different things per reviewer and only 1/3 of the scale actually being used) then going back to the numerical scale feels a bit like them giving up on ever fixing that problem.
For example, instead of stars they could use a bar that fills up, with markers or areas for "decent", "excellent", "masterful", and so on. These areas are unevenly spaced, making it difficult to convert them to a numeric score (and not showing the full bar on screen, e.g. by having the camera pan across the bar while it fills up, would make it impossible to do so). They also have enough room to show if something is not quite excellent, but almost. To add to that, a "masterful" would not necessarily be the absolute maximum, since they can always add markers/areas above that if a later game ends up being even better.
There's another matter that wasn't mentioned here but was in a previous video. I think I remember one of the allies saying it was hard to score a game because they thought it was better than another game that got 4 stars, but not as good as a third game that got 4.5 stars. I think, if they feel that way, that it's better to show that, i.e. literally show at the end of the review "We liked it more than X but not as much as Y".
While unprofessional, I think it would also be a more honest score. Showing the reference scores would do a good job of explaining how good the game is exactly (i.e. some people may see a 7.5 as good and some as bad, but people understand comparisons between the quality of different games, which in the end is what people -- including the reviewers themselves -- actually associate with these scores, no?). In an extreme case I would even use this measure exclusively, because some games are simply incomparable and scoring them on the same scale, whether numerical or otherwise, makes people draw comparisons where they can't really be made. Picking the reference scores from games within the same genre could avoid that.
I'm not saying that my suggestions are good enough (they aren't), but I think it's worth peeking out of the box to see what other possible score systems you could use.