1. The syllabus is asking students to think about repeatability, which is what NESA means by reliability. We want you to learn how to identify and quantify sources of random errors, rather than plugging numbers blindly into a calculator. Anyway, I think it would be extremely rare in Stage 6 Physics and Chemistry to have a sufficiently large number of equivalent results, to be able to calculate a standard deviation. Having said that, I can't see anything in the NESA rules that would forbid such a thing. After all, you are permitted to bring a Casio fx-82 into the exam room, which does stats and linear regression.
2. The accepted literature values are never rounded. I can't think of any reason why you would compromise the precision of something like the Avogadro number, it is just unthinkable. There are a couple of physical constants that are defined to be integers, such as the atomic mass of carbon-12 which is just 12 exactly.
3. Yes, percent error, or p.e. is an acceptable measure of accuracy. P.e. is even more useful when concatenating errors in calculations. However, in Stage 6 as I said we are emphasising the identification and quantification of errors, and categorising them into random and systematic. We don't want to teach formal mathematical theory of error handling before students have a solid grasp of sources of error. To give you an example, if we experimentally determined the enthalpy of combustion and got a percent error of -40%, and we just leave it like that, without probing why it was so low, then we miss the main point of the analysis and discussion.
