I always struggle to understand how 3rd party graders can hope to achieve consistency given that the same person won't always grade a particular denomination of coin, the lighting conditions (whether natural or artificial) will always be slightly different when examining the coin, and it's not possible to make comparisons with other coins of that denomination other than via photographs and we all know how photos of the same coin can differ dramatically depending on conditions and technology used.
Theoretically, the only way to grade truly accurately would be to have a library set of every denomination of coin in every physical (not photographs) grade to make side by side comparisons.
Given that that's not practical, we have to accept that 3rd party graders will have to work from previous photographs of similar coins and written descriptions of grades, much like the rest of us, and can only be accurate to within a range of uncertainty.