Maybe I should try another analogy?
Masterpiece - Treasure - $3+
-----------------------------------
$1
-----------------------------------
When you buy this, you may overpay for it. If you do, gain a Silver per $1 you overpaid.
Adequatepiece - Treasure - $3+
-----------------------------------
$1
-----------------------------------
When you buy this, you may overpay for it by $3. If you do, gain 3 Silvers.
Tolerablepiece - Treasure - $6
-----------------------------------
$1
-----------------------------------
When you buy this, gain 3 Silvers.
Failpiece - Treasure - $3
-----------------------------------
$1
Here we have four cards. MP is the regular Masterpiece. AP is a Masterpiece that is restricted to only overpaying by $3, no more, no less. TP is a Masterpiece where you are forced to always overpay by $3. FP is Masterpiece without overpaying at all.
Should we even bother ranking FP on the $3 list? It is pointless because Silver does everything Failpiece can do better and at the same price.
I think this is the crux of our different views.
I'm saying you can rank MP against $3 cards by considering the extra mount you have to pay to be a limitation in its ability. Or as its strength, depending on which direction you start looking at it form. (If you're thinking, "Sweet I can more Silver!" then it's limiting because there is a cost and you're giving up more expensive cards. If you're thinking, "Man this is a really bad $3 copper" then it's a boon because, hey, it's much better than that.)
You don't like thinking of it that way because, hey, you could have bought a more expensive card for that amount. But I say you just take that into account when you give it a rank. You say that's hard to do, but I say you're really already doing pretty similar tradeoff type estimations when you ranked all the other cards.
I don't think ranking <Fixed>P on the <FixedCost> list is relevant*, because when you're ranking MP (on the $3 card list, of course) you ought to be taking into account its entire functionality---it's the variable and optional overpay that makes the card. It doesn't make it a $6 card, even if you spend $3+$3 on it in 62% of the situations.
If it's going to go on a list other than $3, it has to go on all of them. Otherwise you're picking arbitrarily and inaccurately. Okay, so say you do that, just like you're doing with your <clevername>Pieces. You consider each fixed overpayment and add $3 and keep track in your head that the cost is only $3 for purposes of stuff that cares about cost. You rank it last on the $3 list, obviously. You make up a witty name for the $4 version and rank it there, same for $5 and for TolerablePiece, and for all the >=$7 versions. (How does $6 MP compare to $7 MP compare to $8 MP and beyond? We'll get back to that.) Okay so you've ranked it for all $X>=3.
Now, what do you know about the strength of Masterpiece? Well its effectiveness is a function of the overpay amount, and we've basically listed out the function for each element in its domain. What do you get from looking at one list? Not much, because you're only looking the value of the function for one point in the domain. You're not going to get a complete picture. Okay so you look at all the lists. So, you know it's a bad $3 card, a mediocre $4 card, a good $5 card, a pretty awesome $6+ card (does it fall off? probably. Are $20 MPs relevant in situations other than Feodum?)
(some edits to the end of this paragraph):
Now what does that tell you about Masterpiece over all? Well that depends on where you put it in the lists, but hey you probably have some kind of ballpark "goodness" estimate in your head from all that valuing you did in your thought experiment. But none of those individual lists told us anything about Masterpiece as a whole. We only evaluated its image for each element in its domain, we didn't value the function itself as the object. We should do that and put it somewhere. Where? Well it can't go on a list of any fixed price, because it's not a fixed price. Well, the $6+ list looks promising because it includes cards of different costs. Though really it does this because there are only a few $7 cards and one $8 card and putting them on their own list isn't enlightening enough. But it shouldn't go on the $6+ list because, hey, that's misleading---it can be bought for $3, $4 and $5 and those facts are relevant, or else we'd be looking at a $6 card in the first place and not a $3 card. And also because, on the $6+ list it still has different costs. Is $6 MP better than $7 MP? $8MP? $3+N MP where N is the number of silvers left in the supply? Is a $10MP better than a King's Court? These judgements are now hard because you have variable prices. So then you say, well, you take that variability into account when you rank it. It's on the $6+ list but you have to note that it could be bought for less and it can be bought for more, so you know these things when you give it a rank. But that's my whole point in the first place---the variable cost is what you use to give the card a rank.
So put it on the $3 list, because we've ranked the cards by cost and it costs $3. Yes it has a wonky ability in its overpay, but lots of cards have wonky abilities and that's, well, the entire point of the whole ranking thing, right?
TP is obviously ranked with the $6+ cards. That is simply what it costs.
Now, AP. Now it's getting tough. AP:TP::LGG:IGG, so some of the earlier discussion applies here. But anyway, AP is just FP and TP rolled into one. FP was not worth ranking and TP should clearly be ranked with $6+ cards. So shouldn't AP be ranked with $6+ cards as well? If not, why not?
Finally, MP. MP is AP with a slew more options. Since FP is not worth ranking, then MP shouldn't be ranked at $3 either. But should it go on the $4 list? $5? $6+? Maybe all of them, because MP is different at all those costs, but for simplicity we could just pick the most common one. What is most common? Probably $5 or $6+, because $4 MP is almost as terrible as Failpiece. $6+ encompasses more.
If it helps, I could make analogues for all the other costs too.
The argument for this example comes down to this: if it's not worth buying Masterpiece at $3 in a game, why would we rank Masterpiece at $3 on a list?
Because it is $3; that's what its cost is, and that cost is not an arbitrary number and it does matter. It's also not worth buying Duchess at $2, or Peddler at $8.
*Okay, I really mean "relevant by itself for a given fixed cost". Taking all the costs together and considering the mapping is relevant.