With waveguides, loss-less bends can be made in case the index contrast is large enough compared to the bending radius of the bend. In the optical ray model, the total internal reflection condition is then still satisfied.
1) silicone is a analogue of carbon based polymers - but the chain is of silicon atoms not carbon ones. There are many different silicones.
2) do they deposit the silicone in tracks or are they coating a whole board then etching it? I don't think 45 minute drying time is very practical if you have to deposit it as tracks, but as a board coating it might just work
If the waveguide is not isolated it is probably in touch with the polyimide FPCB. PI has higher RI than silicone. So to form a lossless waveguide during bending in smaller diameter you will need a low RI cladding to the silicone. Fluoropolymer is probably the only choice for silicone. Am I missing something?