If your software tries to anticipate that you might want to not-close when you told it to close, you end up with the opposite complaint: it takes too long for anything to happen.
I'd be willing to bet that there would be far more complaints that your program takes a minute longer to shut down than that they told it to quit when they didn't want it to quit. I'm a huge believer in software that people can use, and love trying to make things totally foolproof for as many people as possible, but I also believe that to truly be useful on a computer you must learn some of the ground rules. Like anywhere else in life. The trick then becomes teaching your users without them realizing or fighting it, not trying to predict the inherently un-predictable (ie, the immediate whims of anyone and everyone).
I'd be willing to bet that there would be far more complaints that your program takes a minute longer to shut down than that they told it to quit when they didn't want it to quit. I'm a huge believer in software that people can use, and love trying to make things totally foolproof for as many people as possible, but I also believe that to truly be useful on a computer you must learn some of the ground rules. Like anywhere else in life. The trick then becomes teaching your users without them realizing or fighting it, not trying to predict the inherently un-predictable (ie, the immediate whims of anyone and everyone).