Could you explain the thumb mimic ratio change in 8ca54d2?

[TannerGilbert]

Hi — thanks for maintaining this repo. I'm using the Revo2 URDFs to drive a simulation/retargeting pipeline and I noticed the recent commit 8ca54d2 ("feat: update thumb joint origins and mimic parameters") made a significant change to the thumb kinematics without an explanation in the message, so I wanted to ask about the reasoning before I propagate it downstream.

For the left hand (the right is mirrored):

-   <joint name="left_thumb_proximal_joint" ...>                                                     
-     <origin rpy="-0.17488 0.23736 -0.16787" ... />   
+   <joint name="left_thumb_proximal_joint" ...>                                                     
+     <origin rpy="-0.33153 0.23736 -0.16787" ... />   <!-- 原点补偿：绕旋转轴 +9deg -->

-   <joint name="left_thumb_distal_joint" ...>         
-     <origin rpy="0 0 0" ... />                                                                     
-     <mimic joint="left_thumb_proximal_joint" multiplier="1" offset="0" />
+   <joint name="left_thumb_distal_joint" ...>                                                       
+     <origin rpy="0.10472 0 0" ... />                 <!-- 原点补偿：绕旋转轴 +6deg -->             
+     <mimic joint="left_thumb_proximal_joint" multiplier="0.691851" offset="0.0395854" />           

Two things changed simultaneously:

1. Rest-pose rotation offsets of +9° on the proximal joint and +6° on the distal joint (accumulated via the rpy x-component).
2. Mimic ratio of the distal joint: distal = 1.0 · proximal + 0 → distal = 0.691851 · proximal + 0.0395854 rad.

The mimic change is the bigger functional difference: previously a distal joint that tracked 1:1 with proximal, now one that tracks ~69% as much with a ~2.27° fixed offset. That materially changesthumb curl geometry and the relationship between the commanded proximal angle and the actual fingertip position.

A few questions:

1. Does this reflect a hardware change (mechanical linkage, cable routing, firmware on the Revo2), or it a calibration correction to more accurately model the existing thumb's tendon behavior that was previously approximated as 1:1?
2. Is the rest-pose +9°/+6° rpy compensation coupled to the mimic change, or independent? i.e. was the previous rpy="0 0 0" distal origin wrong even under the old 1:1 mimic, and these are two separate bundled together, or does the new multiplier rely on the new origin (so the two changes must always be applied as a pair?