You're right, assuming a perfectly efficient market. But, in real life the market isn't perfectly efficient and sometimes an odd property will recover its entire replacement cost.
The argument goes, "If the market will return the full cost of an 1,000 sf of GLA, why aren't there any other properties of this size?" Usually that's valid reasoning, but there could be a couple of good reasons why nobody else has built a large home in that area. The first is that if the markert will return only the costs, but no profit, than a speculator won't build it. The second reason is that perhaps nobody has tried it and tested the market.
We're trying to predict the behavior of buyers and sellers. The best evidence is to find a similar situation, even if it's in another area. The next best evidence is to interview buyers, sellers, investors, and their agents. To categorically state, "If something hasn't been done before, it has no value" could be overly simplistic. Some appraisers are quick to say the extra space will either return its full cost or it will return nothing, but there are some types of super-adquacies that will return a portion of their cost, some with only minimal functional obsolescence.
There was a great joke in the Wall Street Journal a couple of days ago: "Two efficient-market theorists are walking down the street and they spot a one-hundred dollar bill on the sidewalk. They both walk by, knowing that if it were real somebody else would have already picked it up". If real estate markets were completely efficient, AVM's could solve every valuation problem. Human beings make better appraisers because they can provide a degree of common sense to the equation.
The practical answer to your dilemna - keep looking until you find a sale of an equal or slightly larger property. Even if it's fifty miles away or five years back. Then establish a correlation between more recent and/or more proximate sales. That's the only way you can convincingly establish the upper limit of value for your property.
