A claim of someone seeing Bigfoot is most certainly Bayesian evidence. If your friend has been known to be accurate (and his friend, etc.) then it is positive evidence for Bigfoot. (After all, if he claimed to have NOT seen it, it'd be evidence against Bigfoot.) It's just not so strong compared to all the other observations where no positive evidence was found. Ideally you have some perfect way to load up all these pieces of evidence and calculate how probable Bigfoot is. A fantastic example of this kind of work is Gwern's "Who wrote the 'Death Note' script?"[1] Without an authoritative way to experimentally test, it's not following the scientific method. Yet it certainly seems to improve our knowledge.

The lung cancer thing, I'm not sure I follow. Is this simply not statistics issue combined with a lack of knowledge about the human body?

1: http://www.gwern.net/Death%20Note%20script