Dynamic Blending--Possible?

I’m designing a VR game for use with the Valve Index. It’ll need to control each finger individually–so far so good. I was recommended to use a Blendspace 1D to blend between outstretched and curled fingers.

(See the first image.)

However, for those of you who have fingers, you’ll notice that you can curl them in more than just the one way. Look at your index finger-- it can be straight, or it can be touching your thumb-tip like when holding pretend binoculars, or if you keep every other finger straight while trying to bend your index, you get a different curl… I don’t want to cycle between every possible curl on my way from flat-handed to fist.

(See the second image.)

There’s a solution using blend spaces, of course, but this leads to a messy implication. You’d need to circumvent these unnecessary states (pretend binoculars, i.e.) on your way from flat to fist by creating a separate blend space for that specific transition. And so on for n!-1 states.

(See the third image).

For the four states we discussed, we’d have to make six transitions. No good. What’s much better is something like this…

(See the fourth image).

So my question is…how to create this ideal fourth image’s solution?