I cannot speak of Carbon fiber wrapped barrels as I have no experience with them.
I believe that when a barrel manufacturer processes a steel/stainles barrel, he will first gundrill a blank (straight cylinder) to the rough bore dimensions. Then he either cut rifles or button rifles the lands/grooves in the bore. After that, he may or may not lap the barrel, but, he does go on to cut the countour of the barrel. When he cuts the contour, it is common to use a dead center on a backing plate with a dog holding the barrel in place, on the dead center/backing plate. This is done at the headstock of the lathe and the tailstock uses a live center, both ends (live center and dead centers) being inserted into the bore. After cutting the contour with a cutting bit in the lathe, the contour is also belt sanded to remove the tooling marks. It is very possible that the sanding of the exterior of the contoured blank introduces the diametrical errors that the OP's 'Smith observed. With that in mind, I believe it is more correct to say that the contour is not concentric to the bore, not the other way around.
Whenever I thread/chamber a barrel, I always dial in the bore with range rods and .0001" resolution Mitutoyo dial indicators, one on each end. I fix the barrel through the spindle/headstock. I use a 4 jaw chuck in the headstock and a 4 point spider in the outboard (left) side of the headstock/spindle. It takes time to dial in, but I can get concentricity to .0002" to .0003" TIR. I also locate the "high spot" (thickest part) of the muzzle and index it to the bottom (6 O'Clock) position when I cut the threads for the trunion. In doing so, I have isolated the incosistency to the vertical axis, where I can simply dial in more or less elevation on the finished product to compensate. The last place I want that incosistency is at the 3:00 or 9:00 O'clock position(s). If the error is in the 3:00 or 9:00 O'clock position, the more windage correction is required as the target distance increases.
I think there is a common mis-conception that the barrel blank is contoured first, then is drilled/bored. I tend to think that errors much larger than .008" would be very common if the blank was contoured first, then bored/gundrilled. If you think about it, as long as both ends of the bore were prepped such that the live/dead centers were perfectly aligned in the center of the bore (perfectly concentric), the contour could be cut such that the OD of the countoured blank would be perfectly concentric with the OD of the bore.
Personally, I don't get concerned about .002" or .003" of variation of bore to OD of contour. I just index the thick part of the muzzle to the 6:00 O'clock position. Seems to work just fine. What IS most important to me is that the Bore is Straight (unto itself). I really don't care about the bore's relationship to the contour, unless it's ridiculously out of whack.