The term sunk cost has a specific and narrow definition. It's a precise term referring to costs that, independent of the future, cannot be recovered. The sunk cost fallacy thus refers to (irrational) behavior that uses those costs as a factor when deciding something. If the costs cannot be recovered, logic dictates that they not be part of the decision making equation.
It's a simple concept, but one that gets applied far too widely. Say that last night I signed up for a banana of the month club at a cost of 19 cents, billed to me each month. Today I suddenly discover that I have a bizarre allergy to bananas. That first 19 cents is a sunk cost to me since I have paid it and this esoteric club has a no refunds policy. The next 11 payments are also sunk costs since they are guaranteed to happen, but they're not necessarily my sunk costs. If I have a friend, call him... Smeff, who really super likes bananas and he agrees to take my banana club membership then I am now free of the obligation to pay. I've shifted the commitment.
[UPDATE: The example below is wrong. The cost itself is sunk unless I could return the banana to the store.]
Last night I paid 19 cents for a banana. Say that I suddenly discover that I have a bizarre allergy to bananas. Those 19 cents are not (yet) a sunk cost! They were exchanged for a banana, one that I still possess. The banana may no longer have any value to me, but it does to a great many other people. Ergo, some of the initial cost is still possibly recoverable; though the window is time-sensitive so I'd have to act quickly while the banana remains a viable source of eating pleasure to others instead of something squishier.
Two years ago, the Mariners agreed to pay $36 million over four years for Chone Figgins' service. Half of that — the half already paid out to Figgins — is a sunk cost. That cost has already been incurred and cannot be recovered regardless of future events. The other half, the remaining half, is a future cost. There are still ways for a portion of that cost to be transferred to another club.
It is not likely for the Mariners to avoid the full cost of Figgins' remaining $18 million. It's furthermore likely that the Mariners will have to risk getting some sub-optimal performance in the field in attempting to recover some of it. Incidentally and perhaps confusing, the continued playing of Chone Figgins could represent an example of a sunk cost dilemma, but that does not make his 2012 and 2013 salaries sunk costs.
Chone Figgins is not a sunk cost to Seattle. The Mariners electing to start him (should they do so) is not an example of a sunk cost fallacy. If you say either of those things, you are incorrect. Figgins is, perhaps more accurately, a toxic asset; something for which its value has fallen tremendously since its purchase. If you want to argue that trying to recapture some of his future cost is a waste (perhaps by quoting the sunk cost dilemma above), that's legitimate. If you want to say Figgins himself — ignoring the context of who he is a sunk cost to — is a sunk cost, that'd be right too. But that context is everything.