It's common to see statements like "1 Roman denarius at time X was equal to $Y in modern US dollars." The problems with this type of comparison are obvious (all the money in Rome couldn't buy you a toilet, let alone a smartphone). On the other hand, I can see how certain things like food or water are of inherent value in any time or place.
I'm interested in any information about how these sorts of things are calculated and how they should or shouldn't be used. But if I had to narrow it down to a single question:
As a historian, if you're reading about an unfamiliar place or era, what do you do (or avoid doing) with this type of information?
You might find Measuring Worth a useful website. It does a lot more than just give you a number — it gives you a range of numbers and the context to make sense of them. Because as you note, not all of these numbers are meaningful. Even the value of food or water is not necessarily comparable over time easily (the raw price of food has dramatically changed over time, so using it as a "measure" of the value of other things is tricky).
As for what you're using it for, it depends on the kind of question you're trying to answer, and that's what the website emphasizes in its guides/tutorials, and why it is so much better than just a single number. Let me give you an example that is not as far in the past: the cost of the Manhattan Project. The cost of the Manhattan Project in 1945 was $2 billion USD. OK — so how much is that, really? How should we, today, understand that value?
If we look at it as a commodity, then the modern price is something like $28 billion. If we look at it as a cost of labor, then the cost is $58-61 billion depending on the skill of the labor. If we regard it as the economic cost of a large-scale project, then it's more like $183 billion. (There are other measures, I am just choosing three for this example.)
Now which of those makes the most sense to me as a historian? The Manhattan Project was not a massive consumer item so comparing it to bread is kind of meaningless. Labor is not a bad way to view it, because quite a lot of the costs were labor costs (it employed 500,000 people). But since we have an "economy cost" option for a project, that's probably the most sensible comparison for a major government project, and $183 billion sounds about right.
Even then it is pretty remarkable how cheap it feels. That's because after WWII the United States got very used to spending billions of dollars, and now we have several projects whose total costs are in the many hundreds of billions of dollars and even a few whose lifetime costs are in the trillions. That's an artifact of historical shifts in government spending, though, and is something that would be needed to contextualize the number. Another way to think about the number is as a relative percentage; the Manhattan Project was about 1% of the total cost of World War II. That's pretty impressive for a war of that size, considering the Manhattan Project only made two weapons that were used in the war.
Anyway — this is just an example to show that even something relatively recent (in living memory) can be tricky to compare because the underlying conditions are what you are really comparing. These numbers are just a way to understand relative worth, but worth is a tricky concept and needs to be contextualized beyond a simple number a lot of the time. The most useful comparisons with the far-far past, like Ancient Rome, tend to be conversions into historical labor, in my opinion — e.g., if I say to you that something cost X Roman denarius, telling you that equates to some modern USD value is probably less meaningful than saying, "this would be twice the monthly income of most Roman citizens" or something like that.
Again, the website I linked to (which is the gold standard as far as I know for making these kinds of comparisons) explains this all in great detail, and provides lots of information about the methodology.