I agree that they're complementary processes. As far as accuracy, it's less about the care of building as the process itself. A hobbyist mill or router will move the head in the same way, and with the same general degree of resolution, as a hobbyist 3D printer. Where the head meets the work, though, they're spinning precise, carefully ground endmills, while the printer is extruding a bead of molten plastic. It's inherently sloppy, when viewed at the fine scale. Of course, more advanced machines use different processes, and some are much more accurate, but they're very expensive, largely due to patents.
It's a frequently used argument that subtractive processes are wasteful, but in the vast majority of cases, the "wasted" chips are carefully gathered, stored, and sold as scrap to be recycled. Machine shops don't waste metal because they can't afford to, and the majority of metals recycle at near 100% efficiency, with regard to loss.
As additive processes mature, we'll see much more use of them in industry, but it'll be alongside the already very mature subtractive processes that preceded them. Neither is likely to replace casting, for instance. Milling a mold that'll make 100,000 ketchup bottle caps in a week makes far more sense than printing 1000 a week, but printing the prototypes during the design phase is very practical and efficient.