Regarding your calculations about K-code Mustangs, Tony Gregory specifically states in his book that his enumerated total of K-code Mustangs does NOT include Shelbys. At least he says so specifically for the 7,273 K-code Mustangs made in 1965; and, I believe he is implying likewise for the 5,469 made in 1966. If so, your calculations are in error by deducting Shelbys from the totals.
Now I grant you, it seems logical to include Shelbys in the 12,742 K-code Mustangs produced in '65-'66. After all, if I were a junior Ford beancounter in San Jose back then tasked with tallying how many K-code Mustangs went out the door, it wouldn't matter to me whether they went to Shelby or any other dealer. They were still Ford K-code Mustangs until AFTER Shelby got through with them.
But Gregory seems to think otherwise. I'm not sure why, but (don't hold me to this!) it may have to do with the K-code numbers for 1967, where, because the Marti records exist for the 472 K's built, it is known this does NOT include the Shelby K's. So the same distinction is carried back to '65 & '66 (i.e., Shelby and non-Shelby K's are counted separately). Thought I read that someplace, but can't remember where!
As for the ratio of A-code GT's to K-code GT's, computing from data in Jim Smart's "in Search of Mustangs" website, for ALL A's & K's registered there (not just GT's) the numbers work out to 85% A and 15% K. I might expect the ratio (85/15) to be similar for GT's, but lacking more information, I guess it's impossible to know.