Quote - "The default collision mode is cloth vertex against object polygon.
Turn the cylinder into a figure and the constrained group works as expected without any extra vertices."

Ummm... does that mean that the the default collision mode does not have anything to do with the constrained grouping? After all, the long, thin cylinder does have polygons, they just happen to be long and thin....so why would constraining work for long, thin polys on a figure, but not for long thin polys on a prop? Not saying it doesn't, just asking why...

I will remember all this (hopefully). I thought I knew all (errm... most of) the ins'n'outs of the cloth room, but this one caught me out. I'm very happy to now have two solutions to the problem. And many thanks to RedMagickMage for raising the problem in the 1st place..:-))

Cheers,
Diolma

I tried the extra collision options to see if they made any difference (they didn't for a prop), then I just stuck a single bone in and tried with just the default options.
While the simulation was running, I checked what the description was in the manual

"Constrained: Each cloth object can have one constrained group, which is
automatically created with empty contents when you clothify an object. Constrained
vertices “stick” to corresponding polygon faces of the underlying figure."

The following paragraph also mentions body parts.

So you do need a figure (and thus get body parts) before the simulator will do anything with the constrained group.

Cloth vertices in the choreographed group aren't part of a simulation, there completely controlled by the transformation (and morph) dials.