The difference between theoretical and actual yield, as discussed by Michael, is a very interesting issue. The good news is that we can formalize them *both* as resource theories! More precisely, the particular resource theory depends on which transformations you consider feasible. And the latter depends on whether you want to analyze an idealized or a realistic setup, and *which* realistic setup.

For example, if you do industrial chemistry, then indeed the transformation \\(2H_2 + 1O_2 \rightarrow 2H_{2}O\\) should not be considered feasible. In other words, we would have a monoidal poset with \\(2H_2 + 1O_2 \not\leq 2H_{2}O\\) -- although theoretical chemistry would tell us that such a transformation ought to be possible. So what we may be able to get in practice is perhaps something like \\(6H_2 + 3O_2 \not\leq 4H_{2}O\\), which is less efficient. But I agree with Michael that the leftover reagent accumulates, so that we might also have \\(2H_2 + 1O_2 + \mathrm{leftover} \leq 2H_{2}O + \mathrm{leftover} \\). In resource-theoretic parlance, \\(\mathrm{leftover}\\) is a *catalyst*.

@Pierre: I personally haven't thought about it enough to comment on the consequences for economic policy. There are certainly many factors other than maximizing yield to take into account. For example, we may also want a *robust* economy and welfare system, which could be an argument for making it regional rather than global. So I'm not sure, but it's a great question, so perhaps others can say more.