I believe that it is better to play 3/1(2) as it improves your winning chances in term of bear off efficiency. I would not expect to gain a lot by hitting. Although by hitting we "remove" this slot on the 22 point, we also give the chance to white to improve his timming (which is not bad by the way) to "restructure" any deficiencies in his position (which again are not many).
Now, lets see what is the gain in terms of effective pip count in a possible scenario of a successful close out (which is the favorite scenario):
A) After no-hitting play of 3/1(2):
The score (after 0 games) is: a 0, b 0 (match to 1 point, post-Crawford play)
Move number 2: b on roll, cube decision?
| a | 27 | ||
|---|---|---|---|
                     | |||
| b | 25 | ||
| Position ID: 2x0AAHj/CAAAAA Match ID: cAkgAAAAAAAA | |||
• b doubles
B) After the hitting play of 23/22*, 3/1 3/2:
The score (after 0 games) is: a 0, b 0 (match to 1 point, post-Crawford play)
Move number 2: b on roll, cube decision?
| a | 27 | ||
|---|---|---|---|
  | |||
| b | 26 | ||
| Position ID: 2x0AALj/CAAAAA Match ID: cAkgAAAAAAAA | |||
• b doubles
According to GNU in the first case after 3/1(2) blues EPC is 50,38 where in the second scenario after 23/22*, 3/1, 3/2 it becomes 50,79 a total difference of 0,41 pips which
Regarding the RO differences, I suppose they have to do with how XG++ and 3ply can play this position out. It is interesting that the main difference comes from the hitting plays and not from 3/1(2). This means that 3ply can play better this position (after 23/22* 3/1 2/1) than XG++? I would expect the opposite.