My calculation is: $300 difference between FF and APS-C, $300 for larger prism, larger shutter, lager mirror, larger body (compared to 7D). Assuming that the 5DIII have all the features of 7D then, the 5DIII should be $600 above 7D. The current 5DII have less feature than the 7D and sells about $1000 above the 7D. Therefore the 5DII is over priced. Canon should be able to sell the 5DIII at about $2200. Since 5dIII is a FF, Canon may price it at $2800 for deeper profit.
Canon have manufacturing costs (per body produced) but they also have fixed R&D costs. How do they distribute those between the different models ? For example, if the R&D costs average $200 per camera sold, does that mean that entry level powershots "should" cost $200 plus manufacturing costs plus markup ? If you don't want to pay for the R&D, then why are we discussing prices of a model that doesn't even exist yet ? The way it ultimately works is that the way those costs are distributed among their models depends on relative supply and demand for the different models. Higher end models (those where a price increase doesn't hurt sales as much) are likely to cover more of these costs.
In terms of manufacturing costs alone -- Sony has already sold a full frame camera for $2000 (the A850). So it is likely that it is possible to get full frame cameras on the shelves for $2000 a piece with 0 marginal cost per unit to the manufacturer.
But pricing does not depend on marginal production costs alone. It depends partly on fixed costs and partly on supply and demand economics (and these interact as fixed costs are likely to be distributed based on demand). The 5D Mark I has similar full frame hardware to the 5D Mark II, yet it is "underpriced" at $1000. The reason is that there is limited demand for an older model.
A lot of times, people setting the prices are not rocket scientists (surprise surprise). They do not create sophisticated price elasticity models... I implement financial package (ERP) applications in Fortune 500 companies, and it amazes me how rudimentary their cost allocation models are. Most of the time, R&D is considered a period cost and allocated based as a percentage of selling price or product cost, then a margin is added on top of that.
If one went purely with R&D allocation, it would not inflate the prices of the higher end models as much, what does inflate those costs is the % of Mark-up tacked on. Herein comes the art of pricing, the mark-up.
A product manager for lets say a rebel line will work on what he has to work with given his product range and not worry about pricing for the pro series, he will have sales targets (the smarter companies will have profitability targets): These Targets at times are not always set to maximize profitability, but at times are set to mantain market share and customer base till the next model comes along. The real money is made from the higher margin products, therefore for consumers, the higher end models do "rightly" seem over priced, and they are.
I am not sure if the FF's prism is 40% larger that it costs 40% more as well... the cost of material is negligible in this case, the cost over overhead would be about the same, the cost of labor might be the same or even more for the smaller prism (depending how it is manufactured). But I do get that the yields on the sensors follow a different set of rules.
A $500-600 differential between a 7D to 5D class camera (other things being the same) might be justified, the rest is fluff and mark-up!