I checked the calibration of some car tyre gauges today.
I got the following results;
34.5psi - digital gauge (new, from a major tyre manufacturer)
35 psi - balloon (pencil) gauge, (new but cheap)
37.5psi - new digital gauge(*)
37.5 psi - garage type pcl inflator with built-in gauge
38 psi - balloon gauge, (new but cheap)
39 psi - CAT branded Accugage (new, unused)
40 psi - Accugage (not brand new, but little-used, kept for calibration checks...
40 to 41psi - balloon gauges (4-off, old but good quality)
42 psi - CAT branded Accugage (old, pretty beat up)
(*) I got some odd readings with this gauge, see comments below.
I was able to check all the gauges several times on the same car tyre without letting so much air out that the pressure changed. I confirmed this using several of the gauges a second or even third time. I suppose that you would be happy to say that the tyre was probably 39psi +/- 1.5 psi but that still leaves several outliers. I can forgive the Accugages to some extent because they are calibrated at mid-scale (i.e. 80 psi on a 160psi gauges like these, so might be (say) +/- 1.5 psi through much of the range). Car tyre pressures vary anyway with ambient temperature and when running hot; 5psi variations are not ususual because of this. The specified pressures are normally 'cold' pressures. Bicycle tyres normally tend not to run hot in the same way but can still increase in pressure considerably if the rim gets hot during braking.
I tried a few of the gauges on a 50psi bicycle tyre, and if anything, they were in better agreement. Very odd.The balloon gauges mostly only go to 50psi so I couldn't try them at higher pressures.
The digital gauge that appeared to be the more accurate (of the two I tried) in fact gave the most erratic readings of any gauge. However after having used it five times, it appeared to become more consistent. I have no idea why.
I may try all my bike pump gauges yet. The last time I did this IIRC I got over 10psi variation at ~100psi.