The interesting part about the Geller / Voltagestandard products is not that they stuck a reference IC on a PCB, but that they calibrated it with a very accurate bench DMM. For a calibrator, the focus is usually on stability, not initial accuracy. A reference IC might drift less than 0.01% in a year, but the initial accuracy may be only 5%. So you need an accurate DMM to figure out the actual value. Testing all voltage references to be within 0.01 % would be extremely expensive for the manufacturer. Even the very expensive heated zener references (eg. LM399, LTZ1000) are only specified to be 7V ish, +/- 5% or so.