I think Hoov and Twilight are both correct??

Twilight's argument is to use discount model: You should refer to spot rate for delivery of gold at each time period (i.e. the company could sell the gold now for delivery at time 1,2,3 in the future effectively reducing all risk with pricing of gold) and then discount those revenue streams:

$1000/oz for delivery at time 0 X ounces that the mine can produce and deliver at time 0 = time 0 revenue cash flow

$1100/oz for delivery at time 1 X ounces that the mine can produce and deliver at time 1 = time 1 revenue cash flow

$1200/oz for delivery at time 2 X ounces that the mine can produce and deliver at time 2 = time 2 revenue cash flow

These revenue streams should be offset with expected costs (costs including cost inflation) and discounted at a rate near the risk-free rate (note that the company has already locked in the price they get so the risk-free rate is the correct rate to use).

Now, since someone could just buy gold at time 0, sell a futures contract at time 0 for delivery at any of the other times, and sit on the gold until delivery, the difference in price between time 0 gold and the spot has to be the risk-free rate (just financing costs).

Since you are increasing the time 1 price of gold by the risk-free rate and then reducing (discounting) it by the same risk-free rate, you get the time 0 price of gold for all time periods!

Of course, this breaks down somewhat when there is no futures market - i.e. chromite, since a company could not lock in a future price at time 0...