Now that I'm following your thoughts I tend to agree with your 20% starting point. We are both sort of making it up as we go along but I like your reasoning and it's always better safe than sorry. And at only 9% tail area and not that long a moment arm it IS short and small coupled. I was acutally expecting something more like you are suggesting in the first place but reasoned that the short chord distance was overly compensating for the small tail volume.

Once flying SAFELY the balance can always be set using the dive testing method in any event. Or if the model does not have the elevator authourity to actually get airborne then that is another clue that the CG is being set too far forward...

Don't laugh, I saw just that situation one time. A trainer with minimal elevator throw and an overly cautious forward CG based on control line practice. At least it was only funny and the model flew well once the balance point was re-located.