You've done the homework well, flattop1, although research projects I funded for my employer (about a quarter of a million dollars worth over a 10 year period) indicate that the number for use of a torque wrench is often a
bit higher than that and the number for angular turn error is usually a bit lower than those numbers (but not much, in either case). We always expressed ours as the % difference between the highest and the lowest preload in the sample. Angular turn is definitely the most accurate way to establish consistent preload around a bolt circle for the average person IF you have a source for the correct angle for the application (we had a couple of methods available), and ultasonics, if you can afford the equipment, are definitely the way with the best accuracy, but usually cost prohibitive outside of the major production/repair environment because of the price of the equipment, including the necessary shop floor calibration standards.
We had a lot of fun working with Superbolt over the years, both before and after we retired and became a consultant. It a marvelous solution for large fasteners that would otherwise require huge (by engine building standards) torques - we had joints that were in the 1200ft/lb to 4000+ft/lb range for normal torquing, so Superbolt was a big improvement for both new construction and life cycle maintenance.
Jimbo, I have no argument with anything you said, although my comments in my original post assumes that the situation has nothing to do with rusted fasteners. Our experience has been that between the effects of short term relaxation and long term (in service) relaxation, the fastener would move with application of a torque toward the high end of the range specified for the joint. Certainly, if one wishes to loosen the fastener by backing off a quarter turn, there is nothing wrong with adding that step. We will say that marking the joint as you proposed gives an indication that something has changed if they don't line up after retorque, but note that friction coefficient between the threads of the head bolt and the head improves (drops) with each loosening and tightening (you don't want to know how much it cost to find that out), so often what you see with that test will be both an indication of preload loss and reduced friction; that doesn't mean that test isn't proper and useful, just that what is going on can be influenced by more than preload loss due to short term and long term relaxation.
Your emphasis on friction between the threads of the bolt and the head and the bearing surface friction between
the underside of the bolt and the head hits on the most important aspect of establishing preload as consistent as possible among the fasteners in any bolt pattern, not just a cylinder head. Some people get fanatical about buying the most accurate torque wrench they can, spending more than they can afford; it's money well spent, but given the big fastener to faster variation in preload when torquing a joint, they need to know that the portion attributable to torque wrench accuracy is down in the weeds, essentially something they can ignore. Friction coefficient variation from fastener to fastener is the big culprit. On the best day of an experienced technicians life, the difference in preload between the highest preload and the lowest preload in the fasteners in a joint, all tightened to the same torque by the same mechanic using the same wrench, is 33% or worse, 50% variation is not uncommon, and if you throw in a shift change, so you have two mechanics and maybe also 2 wrenches on a joint with 30-50 or more bolts and preload variation can get even higher. With an inexperienced tech who didn't follow the proper tightening procedure closely, we have measured 100% preload variation ( the highest preload being twice the lowest preload) around the bolt circle. But far and away the biggest contributor to preload variation is fastener to fastener friction coefficient variation.
To give a specific example, among many tests we conducted, we did two different friction coefficient tests with alloy steel bolts and within any individual test sequence, all the bolts came from the same manufacturer's lot. The thing is, friction coefficient of fasteners is dependent on fastener surface finish at a level that can take anywhere from 10x to 100x magnification to detect. So we tested fasteners from one specific manufacturer's lot unlubricated and established a high, a low, and a mean friction for unlubed alloy steel fasteners from 1/4 to 1.5 inches. Then during another series of tests we wanted to test in the same size range lubed with motor oil. From a friction coefficient standpoint only, motor oil is not a great lube at the thread psi loads encountered when a joint is fully torqued (the pressure additives in motor oil aren't the type necessary for thread lubrication) but even we were surprised by the results. The fasteners used in the motor oil test were from a different manufacturer's lot than the unlubed test, and while they looked the same to the naked eye, the mean friction coefficient developed for THAT BATCH of alloy steel fasteners lubed with motor oil was actually slightly HIGHER than in our previous test of unlubed fasteners.
So Jimbo's emphasis on friction cannot be overstated - it is the single biggest factor by far in determing actual preload with application of a specific torque.
Still didn't write a book, but this is the start of a chapter, lol.