Look at the circuits for DC error sources:
1. input resistive divider (50-100 ppm/T),
2. buffer amp (AZ),
3. second amp interfacing.
The author solved a "zero set" error at 2, but introduced it back at 3, since minimum output voltage for single power is newer zero.
There are also many different problem, like full scale error, drift in the ADC itself and resistors, that AZ opa egotistically doesn't care. So what is the point to spend money on AZ, if overall accuracy is no difference to low cost INA?
I would put a reed / opto relay right at the front to re-calibrate periodically , though price for relay could make this approach less cost effective.