Well, traditionally (if something that's only been around for about 50 years can be called "traditional"), when using multitrack tape, each track of the tape holds one track of audio, of course, by definition. Often some parts are doubled by using two mics, say, and using two channels of the tape. In mixing, each channel is one sound, and you would pan it in the stereo spectrum as you wanted, and finally record a two-track master off your mix.
With computer-based recording, you have the option of recording a track as either a "stereo" track (using two channels of input, often physically connected into a single stereo jack on cheaper soundcards and therefore can be easily thought of as a single stereo track) -- that is, you assign both L and R side of the input jack's signal to one sequencer track at the same time.
Or you can assign each side of the jack's signal to a separate track. Then you record the same thing but they are two distinct mono sequencer tracks.
The reason the latter way is better is control. Once you have the sound recorded into a single sequencer track, you can typically only pan the two tracks together, or raise and lower the volume of both together, or add plug-in effects to both together... in a sense, you have sort of pre-mixed already, and it fixes some of the possibiliities you have.
If everything is recorded to a mono track, you control each track completely, individually, in the standard old traditional way. Your panning assignments in the mix affect what happens in the stero image, captured in your stereo master.
(Actually, after saying all that I realize it might not be true; I know in Sound Forge I am able to play with each half of a single stereo file independently... I'm not sure if I can do that in Cakewalk or not. If so, I guess it's six of one and a half-dozen of another. Then the only real reason would be consistency and less potential for confusion. I like to think of each part seperately, it's just much clearer in my mind.)
