We are pharmaceutical packaging plant and we use now about 450 colors from Pantone guide (400 Coated and 50 Uncoated). We have an agreement with ink supplier, who make for us specifications with L*a*b* values and colored sample (it's special tool where this samples are made – I don't know the name of this tool) for each color. Then I send this L*a*b* values to our suppliers (printing houses – we have about ten of them) who are obligated to print each color within dE=8. We use delta E CIE L*a*b* (formulae CIE1976). I have made some test (about 50 different colors along whole spectrum) with formulae dE2000 but I didn't percieve some special benefits of this formulae. We all (our company, ink supplier and printing houses) use same instrument (X-Rite SpectroEye, D65, 45/0, 2°, POL filter).
But … I must admit that we have some problems. Colors are very alive. That means suitable color (within dE=2-3) right after printing process, but unsuitable color after day or two (or maybe a week) when ink penetrates into printing material (especialy Uncoated papers). Some colors (red, dark blue-Reflex Blue and violet colors especialy) have a predisposition to vary with time (sometimes more than dE=12).
So we are trying now to determinate those very problematic colors (less than 10% of all colors). I think that we will determine »higher« dE for those colors or maybe just measure new L*a*b* reference of dry color.
I partialy agree with Gordon (don't believe there is a delta E tolerance for PMS colors. Even the latest Pantone GOE library does not mention tolerances or CIEL*a*b* values.
Typically swatches/draw downs are acompanied with a "Hi-Lo" density reference. BTW a Delta E of 5 can be a good match as the visual perception of difference depends on the hue you're measuring and how the Delta is being calculated.) but we recieve more than 100 packaging materials a day (folding boxes, leaflets, leabels, foils, ….) so we have to have suitable (fast) system for estimating colors.