The 1996 to 2000 period was remarkable as the dot-com IPO crazy period. It's the closest, and most optimistic investment environment to what we are seeing today, thus makes a good comparison for making an optimistic valuation model for Facebook.
I believe puzzle questions, made famous by microsoft, and then copied by google, are the exemplar here. The standard back-story is that it is not about getting the right answer, but explaining your thought process. Unfortunately, the act of explaining totally wrecks any deep or involved thought process, because you are constantly engaging the speech parts of your brain, which upsets your working memory. Your interviewer never had to go through this process while talking out loud, so has no idea what the question is like on the receiving end.