I'm not sure that you're considering all the options.

1. Abandon A. Net profit from now $Y (sale price of building for A).
2. Replace A with B. Net profit dependent on NPV of B's return.

So you may not subtract $X from B, but you do compare NPV(B) to $Y.

Also, remember that we're not trying to figure out the return on B. We're trying to figure out what to do next. In terms of planning, the sunk cost is irrelevant. But in terms of calculating return, A was a turkey. And to calculate the return, we would include $X in our costs for B. And for the second option, we'd subtract $X from $Y (may be negative).

Sunk costs are irrelevant to planning, but they are very relevant to retrospective analysis. Please don't confuse the two. When looking back, part of the cost for B will be that $X. But in the middle, after paying $X and before starting B, the $X is gone. You only have the building and have to make your decision based on the options you have at that moment.

You will sometimes hear $Y called the opportunity cost of B. You could sell out for $Y or you could do B. You should only do B if it is worth more than $Y.

The sunk cost fallacy would be comparing B to $X. Assuming $Y is less than $X, this would make you not do B when it is your best path forward from that moment. I.e. $Y < NPV(B) < $X means that you should do the project. You will lose money (apparently that's a foregone conclusion), but you will lose less money than if you just sold out.

You should also do B if $Y < $X < NPV(B) or $X < $Y < NPV(B). In general, you should do B any time $Y < NPV(B). The only time you should not do B is if NPV(B) < $Y. If they are exactly equal, then it doesn't matter financially whether you do B or not.

I'm not sure that you're considering all the options.

1. Abandon A. Net profit from now $Y (sale price of building for A).
2. Replace A with B. Net profit dependent on NPV of B's return.

So you may not subtract $X from B, but you do compare NPV(B) to $Y.

Also, remember that we're not trying to figure out the return on B. We're trying to figure out what to do next. In terms of planning, the sunk cost is irrelevant. But in terms of calculating return, A was a turkey. And to calculate the return, we would include $X in our costs for B. And for the second option, we'd subtract $X from $Y (may be negative).

Sunk costs are irrelevant to planning, but they are very relevant to retrospective analysis. Please don't confuse the two. When looking back, part of the cost for B will be that $X. But in the middle, after paying $X and before starting B, the $X is gone. You only have the building and have to make your decision based on the options you have at that moment.

You will sometimes hear $Y called the opportunity cost of B. You could sell out for $Y or you could do B. You should only do B if it is worth more than $Y.

The sunk cost fallacy would be comparing B to $X. Assuming $Y is less than $X, this would make you not do B when it is your best path forward from that moment. I.e. $Y < NPV(B) < $X means that you should do the project. You will lose money (apparently that's a foregone conclusion), but you will lose less money than if you just sold out.

You should also do B if $Y < $X < NPV(B) or $X < $Y < NPV(B). In general, you should do B any time $Y < NPV(B). The only time you should not do B is if NPV(B) < $Y. If they are exactly equal, then it doesn't matter financially whether you do B or not.