How you do it depends on what you want to achieve, really. With both of our tastings it was to test how well people really knew beers they were familiar with. For this you need to say what the beers are in advance and I agree that five is about the limit. If you have someone who insists Beer X is better than Beers Y and Z of the same style then they're asking for one of these.
I think the totally blind rating thing is a great idea. It'd work best outside the pub context, I think, where the person, or people, setting it up have a greater choice of candidates available. Because there's a higher level of secrecy involved, the person who bought the beers can't do a tasting based on them, because their knowledge of what's there will skew the results. The purest way would probably be to have the organiser not taste, but that's no fun. The alternative is to get a large number of people to bring six or so random beers of varying quality, let everyone see the beers available at the start, and then give each taster only a small sample of them.