Oh boy... you'll probably get a lot of long responses on this. There are a number of reasons:
1. You already alluded to it - many dendrobatids are rare in the wild, and though people argue endlessly about the virtues and realitites of "captive maintenance for insurance against extinction," there is inarguably at least some theoretical merit to this goal.
2. Many, though not all, hybrids of other herps (colubrids, diamondxjungle carpet pythons, etc) are (I believe) naturally occurring integrades. This does not appear to be the case with darts, many of which are geographically, and hence genetically, isolated. Thus, dart hybridization is an unnatural event.
3. As a follow-up on #2, sometimes hybrids don't look like hybrids, but rather seem to have all the characteristics of one parent or another. Breeding this hybrid to a "pure bred" unknowingly may introduce genes which, down the line, can pop up and affect future generations in a variety of ways, from throwing out odd-looking offspring to, at least in theory, creating homozygous recessive combinations which lead to sickly, deformed or otherwise diseased offspring. edit: the disease argument here is kinda weak, as this can happen with simple inbreeding too, but the "weird offspring down the line" argument has more merit.
4. I think there is a strong emphasis/desire in this hobby to keep things "natural." Everyone wants a naturalistic vivarium (you basically never see weird gravel colors, for instance). That applies to the frogs, too. These things are so damn beautiful in their natural colors that people are almost offended that anyone would want to try and modify them.
5. Many, myself included, are dismayed at the thought that darts could go the way of the ball python - every little phenotypic difference, from the significant to the minute, is hyped as some great new morph commanding thousands more dollars. You want a laugh? Look at
http://www.kingsnake.com classifieds for ball pythons; some of the animals there are CLEARLy just nice normal balls, but they are advertised at outrageous prices. Yellow-belly, I think, is the biggest ripoff joke in the world... but I digress.
6. We still have a tremendous amount to learn about within-species variability for almost every dart frog species. The incredible number o naturally ocurring pumilio, auratus, imitator, etc morphs is bewildering, and figuring out what the natural localities are (and what they look like) will go a long ways towards helping our understanding of these animals and their status in the wild. Creating (and spreading) hybrids could irrevocably damage this knowledge, even as isolated populations begin (or continue) to disappear.
There are other reasons, but I think that's a good start.