So the absolute resistance depends on the doping, which is tricky enough to control (furnaces, impurities, voltages, sputtering, etc.) that even with good contr

So the absolute resistance depends on the doping, which is tricky enough to control (furnaces, impurities, voltages, sputtering, etc.) that even with good control, the final resistance is within 20-30% of the target. The relative resistance comes from the shapes themselves: it's the same material for both resistors, with the same properties from the same doping. And photolithography is great at matching shapes and patterns, so you get fantastic relative tolerancing. There are, of course, still plenty of other sources of error.