I think you're confusing the tool-calibration step with the actual measuring. As long as a scale is built symmetrically, an empty scale will read as balanced in any environment. Both sides are made from the same materials with the same densities and displacements etc.
Once you apply the dissimilar samples (a known 1kg of styrofoam and a known 1kg of steel) it will cease to read as perfectly even, because the air-buoyancy of the samples will be different. Perhaps not enough to see, but it's there.
Then you submerge the system, and the difference becomes obvious.
Ah, you seem to be assuming that the mass of the crown is known and an equivalent mass of gold can be produced.
> What we're actually trying to check is that the readout doesn't change ...
I'm not sure what you mean by readout. If one side is A and the other B the balance only has three readings, A>B or A=B or A<B.
Typically I would expect the crown to be put on one side of a balance. Then the weigher would search for the amount of gold that makes the balance, well, balance (A=B). In effect the weigher chooses diff_air==0.
> A good counterfeiter will ensure diff_air==0
I don't see how the counterfeiter has any influence on diff_air, since the mass is not known beforehand.
Your equations seem correct.
The first is effectively true because you can weigh the crown in air, which for goldish-density crownlike objects is so relatively thin that you'll get similar results to doing the comparisons in vacuum.
The second assumption of a reference-sample was explicitly stated earlier in the thread, inside the Wikipedia quote.
> I don't see how the counterfeiter has any influence on diff_air, since the mass is not known beforehand.
The crooked crown-maker already knows (A) how much true-gold mass their customer is expecting them to deliver and (B) they have many opportunities to mass-measure and adjust the profitably-adulterated not-quite-pure-gold object they are creating.
The trick being described has the weigher balance the scale in air, with unknown mass A and some mass of gold B. The two sides are equal in weight in air, different from mass. Then the entire balance system is submerged. The difference in density of A and B, and therefor volume, leads the balance to become unbalanced.
It would be wise to add mass to the lighter side to try and measure how imbalanced. I would bet most adulterants are about half the density of gold or less. In g/cm3: Gold is 19.3, silver 10.5, lead 11.3, copper 9.0, nickel 8.9.