I once attended a talk by someone working for Amadeus, a software company that specialises in software for the travel industry. ("Selected Challenges from Distribution and Commerce in the Airline and Travel Industry", by Francois Laburthe, CPAIOR 2008)
Unfortunately, I seem to have lost my notes, but from what I remember, a key point was that the reservation systems not only calculate with the known fare classes, but subdivide these classes further many times over. Each of these subclasses has its own price tag and may only be available within a given time window, i.e., a certain number of ticket for a given class and a given price may be available so many days before the flight between hours x and y (yes, the price may depend on the time of day). Whether a certain price is quoted to you then depends on, among other things, the time the request was made, whether the system thinks you're a business traveler or not, whether the system thinks that it will be able to sell seats in this class at a higher price at some later point, etc.
Here are two excerpts from the talk's abstract:
Airlines are increasingly moving towards complex commercial policies, with numerous fare classes potentially available on a given flight and where the decision of the actual availability of a fare depends on many characteristics of the request (end-to-end journey, channel through which the request is arriving, membership to loyalty program tier, ...).
and:
Revenue management systems include a forecaster (estimating the future potential sales of tickets from now till the day of departure) and an optimizer (defining the appropriate inventory controls from the forecasted demand). Modern airline revenue management systems include end-to-end availability logic which states that an itinerary should be available for sale if its yield to the airline is greater than the sum of the opportunity costs on each segment of the trip. Moreover, this network logic is often furter tailored by means of fare modifiers which further restrict the availability in cases of hints that the request is originating from a business traveler.
Note the statement about "end-to-end availability logic". That means that the longer trip will be available at a lower price if the expected yield is higher than what the system expects to earn from selling tickets for the individual segments.