So I talk about the pose reader v1 and v2 in the scripts section of my site, so I won't cover them here. This isn't really a new approach, it's more of a tweak, hence v2.5 instead of v3. So the issue with the sphere as I was using it was that it could only handle four poses. This was intentional because I always did four corrective poses, but as I have continued to push the quality of my deformations I found that I needed more poses to blend between in order to get the desired results. Below is a video of a WIP rig which shows of the issues.
The solution was to introduce more poses, so I altered the script to allow for any number of poses, in the case of the shoulder 8 poses.
The tradeoff is that you need to all the sculpts to get desirably results. I'd like to figure out a way to use different textures to get other more user-defined results. I am still trying to work around an apparent bug in Maya that makes using the pose reader unintuitive. It seems as though the layered texture node cannot preview the result of 8 color inputs. As you can see below the shader matches the render, but the view port is very different (the layered texture node is effecting all downstream nodes). I have tested this with 6 and 7 zone and those work as expected 8 or above however results in this issue. It may finally be time to move onto Maya 2017.
Another problem (user error in this case) I came across was that by taking the output weight and plugging it into a animCurve I was adjusting the return weight for a whole zone rather than just adjusting the rate of vertical falloff. To solve this I used remap value nodes for just the null zone ramp of each zone in the node network. These nodes are not created with the reader by default.
Not quite sure what a Pose Reader V3 would look like, maybe incorporating translated and rotates? I think I'll just have to wait until a project pushes V2.5 to it's breaking point and re-evaluate from there.