The rigidity buys you a lot. It's not just easier modelling, it's also higher reliability. If your bond is rigid, you don't have to worry too much about heat causing errors in the synthesis.
Yes, there are interesting designs with biopolymers, but they probably cannot get you the type of atomically precise manufacturing that you need for most "cool" MNT applications.
The thing to recognize is that tentacles predate arms, legs, or fins by millennia. Jellyfish tentacles have fewer kinds of parts than any arm, hence they constitute simpler machines.
In the paper you cite, humans are describing cephalopods with a model based on their own anatomy; it is not the other way around, as you claim.