I have absolutely no scientific basis for this, it's simply my own observation, but from my days as an undergrad it seemed like I can get the same amount of "dreaming" (as measured in "perceived length of the dream") done off of a 20 minute nap in class as from a good 6-8 hours on the weekend. So, if sleep time doesn't have any effect on dream time, then it makes sense that any dreams we "remember" are simply the random creations we come up with at that brief moment of waking up. And if you'll permit a computer science analogy, it seems to me that dreams are just what happens when you read uninitialized memory after turning the power on. In other words, our brains are off in crazy rest/sleep mode, and switching back to awake mode leaves a lot of stuff all garbled, and we interpret that as dreams. If that's true, then it makes sense that it's hard to remember because those "experiences" aren't attached to anything, just random thoughts. And there's all sorts of information about memory being enhanced (if not entirely based) on being attached relations to other ideas/sensory experiences. So that's my guess. Sorry it's all just conjecture though...
(And by the same token, that probably supports "traumatic bad dreams" sticking with us, because we immediately associate them with the traumatic ideas/memories we have in real life, causing them to stick.)
Anyway, I'm curious to see if this is at all supported by research out there =P