A bunch of things apply.
They have to pick a mission success criteria and they will pick a reasonably softball one. They have nothing to gain from setting the mission success criteria to be more aggressive than it needs to be and lots of incentive to make it an easy target to hedge against unforeseen circumstances.
There is always quite a bit of uncertainty in missions like this. Spirit and Oppy were the first big, independent rovers to go to Mars and while NASA simulated and prepared for the mission environment, in the end they couldn't be sure that everything would work as intended. You can see this in the faulty wheels that Curiosity has - despite three previous rovers worth of experience, NASA still underestimated the wheel erosion that their larger, nuclear-powered rover would experience. NASA has adapted the way they drive it to compensate but the problem was very serious and quite a concern when it was noticed.
Even when everything goes right, certain parts and instruments have limited life spans. The drills that Oppy and Spirit used to sample rocks were a good example of this as it was known that the bristles they used to chew into rocks would wear out. So if you have instruments and parts with limited life spans, you are inherently putting a timer on success that can only get worse, not really any better.
Let's say you have two instruments, a chemical sensor with a theoretically infinite life span and a drill with a life span of 45 days of heavy use. Now when you get to Mars, you find that due to an unknown design flaw, the chemical sensor dies. This means you have a useful scientific life of 45 days for your rover. If you have tied your mission success criteria to say, 180 days of useful scientific observations, you will all of a sudden find your hardware struggling to meet a goal that the supposedly immortal chemical sensor could have easily met. You want to avoid boxing your mission in unfavorably and provide plenty of margin for success. Under promise and over deliver.
Another big issue is budgeting. Big flagship missions are expensive both upfront and on an ongoing basis due to the teams of engineers and scientists you have to pay to continue flying/driving them. Because of this and the way our flaky, year-over-year spending bills are handled by Congress, NASA can't always budget all of the money they need for operating the mission long-term. They choose instead to rob Peter to pay Paul by taking money out of the future ongoing-operations budget to pay for the immediate construction and launch of the mission - or rather they construct the budget such that it's heavily front-loaded to begin with, even if it means they can only afford a short duration of operation. If you can claim success early into the life of the mission, it's easier to go back to Congress and ask more money to fund continuing operations than if you go back to them without total success already behind you.
Local conditions also play a role as well. The recent ESA mission called Rosetta is a good example of this in that the orbit of the comet it rendezvoused with was known to be nearly impossible to survive long-term. Similarly with Mars, yearly dust storms can range from very mild or even non-existent some years to absolute beasts that black out the sun for months. Given these kinds of uncertainties, you really want to make sure you get mission success fairly quickly so you design mission goals that can be accomplished in a short amount of time to avoid hazards. After that you can begin exploring more long-term investigations.