I guess the assumstion is that your shutter and light meter are within tolerance.
Dear Paul,
Yes, I've been thinking some more about this, and the more I think sbout it, the more it looks like the flexibility of the process being used to plaster over potential gaps.
Camera+lens flare factors for the cameras likely to be used will vary from near unity to 2x or 3x.
Meter variations of +/- 1/3 stop are more usual than not.
People often read the same meter, with the same indicated exposure, either optimistically (rounding up to the nearest 1/3 stop) or pessimistically (rounding down to the nearest 1/3 stop).
Shutter speeds are often 1/3 stop or more slow, and efficiency varies with aperture -- though with a very short focal length in a normal shutter, efficiency can be so high as to result in relative underexposure. I find this with my wife's 35/5.6 Rodenstock Apo-Grandagon.
Now, I have little doubt that the tests you can 'buy in' will remove several potential areas of uncertainty, and will therefore result in more certainty and more precise exposure. I have equally little doubt that often, the potential errors that are not tested for will either cancel out, or can be estimated with enough accuracy to increase certainty still further.
But I am equally confident that in a worst-case scenario, with all the untested errors cumulative, an error of a stop is extremely likely and an error of two stops is possible.
As I say, this doesn't matter very much: the flexibility of B+W neg/pos photography will be able to swallow it, especially if the testing procedure is biased towards a lower EI than the true ISO. But it does argue that some people may think they are working to rather more precision than they are.
To be thoroughly cynical, given how many people see what they want to see (especially if they have paid a lot of money to see it), a 'cowboy' operation could probably pretend to be doing the same thing; make up sets of believable figures; and still rely on around 50 per cent satisfied customers.
I'm not saying for an instant that this is what is happening: I'm sure that the testing procedure from an APUG sponsor would be as good as it could reasonably be made. But I do agree with the poster who said that unless you're prepared to learn a bit of theory, and do a bit of testing for yourself, you might be as well off with the Massive Development Chart.
Cheers,
R.