Well, you probably are right that we're not really totally consistent, at least in our reasoning.

I was just noting that we as individuals kind of need to resolve those inconsistencies. Generally, when you notice an inconsistency in your preferences, you're able to either decide you prefer one to the other or say you don't know, at which point you (at least for the moment) assign them equal rank, right?

Further, I was noting that even if for some stuff you're not entirely clear on, there're cases where you definitely prefer one thing to the other all else being equal. (The case of not being equal is where you start computing the knock-on effects/separate consequences and modify the net value based on those modifiers...)

Then from there, one can construct utilities by looking at differences between how much stuff is valued. (Which basically can be done by going "if I'd be equally surprised if A or B occur, if I have no particular expectation one way or the other... And if I'm offered to either conditionally replace A with C... that is, if A would have occurred, instead C will occur) or to replace B with D, then whichever one I choose, I'd assign a higher difference to.(ie, utility(C) - utility(A) > utility(D) - utility(B) if you choose the first option, or if you're indifferent, you set those differences as equal.) From there you have an internal "currency" to use, and you can pretty much dutch book argument your way up to Bayesian probability (and the rest of decision theory, I believe). (Well, okay, to actually make the Dutch Book argument work here, we have to focus more on your potential losses and gains rather than the opponents, and more "how much potential utility are you willing to sacrifice to gain some other amount of utility under some circumstances", but that's more just rephrasing and such)