The point of them is to show you can creatively use what you know to come up with a very quick ballpark figure. This sort of estimation comes up surprisingly often in real life:
1. In physics labs, it was often useful to do a quick estimate of roughly what the answer should be and the likely direction of the error. That way, if we got something way off, we could check the equipment immediately and rerun the experiment if necessary, instead of going home, analyzing everything, finding out we were way off, and having to go back to the lab and collect everything again.
2. When selecting an algorithm, it's good to be able to estimate roughly what the input size will be, so you know if say an N^4 or N! algorithm will work, or if you really need to get it down to linear or N log N.
3. When selecting a startup to work at, it's useful to do a back-of-the-envelope estimate of its market size, to give you an idea of what the startup may sell or IPO for and hence how much you can make off stock options.
4. When founding a startup, it's good to do a back-of-the-envelope estimate of your costs and potential revenues, to make sure that you're not totally uneconomical. (This sort of estimation could probably have prevented debacles like AllAdvantage or Pets.com.)
5. When investing in growth stocks, you can get a good sanity check by estimating their potential market size. For example, in 2002 Akamai was trading at a total market cap of $150M. I did a quick calculation based on the number of computers on the Internet, and figured that the only way they could conceivably be worth $150M or less was if they went bankrupt. Then the question becomes the strength of their product offering and their capital structure, and both of these had fairly easy answers.
http://www.joelonsoftware.com/articles/fog0000000073.html
(Ctrl-F impossible question)
Regarding not recognizing your own ignorance, I think you definitely have to mention in your answer that really you are making up all the numbers and just tackling it in the first logical way you came up with.