As an anonymous poster mentioned in yesterday's comments, there are some differences between houses when it comes to the acquisition P&L, so some of the examples I'm about to cite may not hold for all publishers or editors in the business. Hopefully they'll still illustrate some of the strange nuances of the acquisition process.
So, as I mentioned yesterday, one of the exceptions to the book-must-make-profit rule is in the case of books that for whatever reason are expected to sell x copies, but with additional marketing dollars and sufficient support from the house may sell many multiples of x copies. As anon noted, some analyses incorporate these numbers into the acquisition P&L from the get-go, and some (like the P&L system with which I'm familiar) don't do this, but will instead include the increase in net sales and marketing dollars in one of those previously mentioned intermediary P&Ls.
Depending on how the P&L is run, this may or may not constitute an exception, but I think it's important to note that a P&L can reflect a loss for a potential title under one marketing scenario and a profit under another.
Another occasional exception to the rule is the short story collection. As I've mentioned before, short story collections don't really sell, but every once in awhile you see a house put out a collection of short stories by some hot-shot up-and-coming author with an Iowa MFA and ninety-three minor literary awards and publications in Very Serious Literary Journals. Why does the house publish these collections if they don't sell well?
Well, oftentimes said up-and-coming writer will be working on a novel—a MEGA SWEET novel—but since it's not done, all they've got to sell is the short story collection. A publishing house, in order to secure rights to the (potentially very lucrative) novel, will buy the short story collection on the condition that they also get the novel. The publisher then takes a loss on the short stories and hopes to earn their money back (and then some) on the novel sales a couple of years later. And that's something I can get behind: the theory that one invests in authors, rather than solely in individual books.
A related exception is the multiple book deal. Once in awhile, a publisher may project a slight loss on a book, but not on a two-book deal, hardcover + trade paperback combination, trilogy, &c. Arguably, since the only P&L that matters would be the analyses run on the set of books (and not the first book alone), this doesn't really constitute an exception, but I think it nonetheless explains why the first book by an author may lose a bit of money or break even while being fairly in-line with initial projections. The goal is to make money in the long run, not necessarily in the first year of the very first book.
In fact, I find this so important that I'll say it again: a publisher's investment in books, much like an individual's investment in the stock market, is most effective when done with an eye for the long-term outcome. Myopic projections and money-now-consequences-later sales strategies may seem great in the short term (oh, how I love puns), but without solid backlist sales—which account for a staggering amount of billing—and steady movement of inventory, you're pretty much sunk.