First, XG's opening book has making the two point (P) better than splitting (S) after 65R-64 by .012, a lot more clear than GNU's rollout.
GNU's rollout was 31k trials using 2 ply/2 ply (XG 3 ply) for both cubes and checker. This was the standard default solid GNU settings at the time.
XG's opening book rollout was only 10k trials however stronger settings of 4 ply checker and XGR+ cube. There is hardly any variance left in either rollout.
To see the difference between XG's and GNU's strength of these rollout you can check out the ply comparison provided in the link.
Your own XG rollout at 7a 7a is only 3ply/3ply which takes a significant hit on cubes. There is no real reason to change from the default of 3ply/XGR in my mind if you can perform a ~700 game rollout 3/3 in a couple of minutes from an early game position. Your rollout shows a surprisingly large difference between the plays, .048! Being only 700 games though there's a lot of variance left and due to that I decided to toss on my own rollout shown below which is much more in line with what we'd expect. (2k trials, 3ply/XGR) Yes, your rollout has the search interval set to large which I think is unnecessary as it adds very little strength wise to normal positions and unnecessarily wastes CPU time.
                      |
| ||||
| XGID=-a----E-C---fE---c-e----B-:0:0:1:64:0:0:0:7:10 | |||||
| Blue to play 64 | |||||
| 1. | Rollout1 | 8/2 6/2 | eq: -0.182 | |||
| ||||||
| 2. | Rollout1 | 24/18 13/9 | eq: -0.192 (-0.010) | |||
| ||||||
| 1 2592 Games rolled with Variance Reduction. Dice Seed: 73253323 Moves: 3-ply, cube decisions: XG Roller | ||||||
eXtreme Gammon Version: 2.10, MET: Kazaross XG2
As to your original question 'I can't figure out why burying two checkers early on can be the right move. Please explain.' part of the problem is how you view making the two point. These checkers that are used to make the two point in this position are anything but buried or dead. The terms buried or dead imply that they won't serve much of a purpose. Instances where I might consider such checkers as buried or dead is when the opponent has a higher inner board anchor or even the potential to make one on his upcoming turn. Maybe we have a dilly builder on a point where that dilly can only go behind our opponent's anchor. General examples of what I just said in writing below.
      |
| ||||
| XGID=--B-b-D-B-----------------:0:0:1:00:0:0:0:7:10 | |||||
| Blue on roll, cube action? | |||||
eXtreme Gammon Version: 2.10
 |
| ||||
| XGID=-aBa--DaB-----------------:0:0:1:00:0:0:0:7:10 | |||||
| Blue on roll, cube action? | |||||
eXtreme Gammon Version: 2.10
  |
| ||||
| XGID=---bC-D-B-----------------:0:0:1:00:0:0:0:7:10 | |||||
| Blue on roll, cube action? | |||||
eXtreme Gammon Version: 2.10
Making the two point gives us another point in our home board therefore giving us the stronger board for the time being. It does not take away our ability to also play a racing game. Making this point after 65R-64 makes some numbers a little more awkward to play than if we had split such as [51 41 21 55] That is of course only in the immediate sequence. On down the line should we hit something we'll be more than happy to have the extra point made increasing our opp's fanning numbers from one to four, but also making other entering numbers once again awkward such as [52 42 53 etc..]
I'm not going to try to explain the tiny difference between the plays because that's impossible to pinpoint. We should however be able to get the big picture of each different game plan and how that game unfolds with Play A v. Play B. Right now I'm guessing that your relationship to making the two point is unjust and you should practice by playing it out or even by playing 64P as the opening roll as I assume you normally don't play that either and it is a perfectly valid option.
Stick