>>any difference in the outcome would imply an
>>inaccuracy in the simulation.
>Not correct. Think of my example of a written
>description of a hydrogen atom. The description
>supposedly contains all the information about
>the atom that exists, yet the piece of paper
>with marks on it is not a hydrogen atom.
I don't get it. Are you stating that the product of a hydrogen atom is simply "being a hydrogen atom" and has nothing to do with its responses to stimuli? I would say that is not the outcome at all but the source thereof. Certain energy organized in a certain way produces these results when interacted with in certain ways -- in all essential respects, that is what a hydrogen atom is.
I notice that your example omits a computing machine that actually reads the paper and uses it as a basis for responses to other information inputs. I would consider that a crucial aspect. Otherwise it's about like debating the vital status of a hypothesized perfectly cryopreserved patient which is never awakened. One could say they don't die until they are destroyed, or one could argue that they died the moment they lost consciousness -- but it would of course be a pointless terminology debate about the meaning of the particular word "death", not proving a single thing about the actual universe. Similarly your paper example blurs the line of "real" in a pointless and confusing way by taking away the actual processing of the information, which is an important part of what we mean by "real" in the context of "hydrogen atom".
Ultimately, saying that the paper with a complete programmatic description (combined, presumably at some point with a turing machine which simulates it) is not a hydrogen atom is simply begging the question. How do you know it is not a hydrogen atom? What property is essential to our definition of hydrogen atoms which this lacks?
In your earlier example, you mentioned that an uploaded human cannot drink non-simulated water. You haven't justified why this should make a difference -- rather you responded with the preemptive insult that this is a poor simulation of an argument. But the claim that an upload is not equal (in important respects) to a physical counterpart is what we are trying to evaluate -- why should simulated water be less important than physical water for this argument?