Unfortunately, I think you have sketched out the correct workflow. There is no auto-magic way of doing this.
You still have to reference a series of locations on the scanned map image, with its real world location (your county boundary corners or whatever).
It would be difficult to automate because the scanning process could not guarantee the exactly same pixel coordinate for a GCP everytime. (You might have put the map through the scanner at a slightly different angle perhaps etc.)
