When it's using an exploratory methodology. Require a falsified null hypothesis before conducting every observation and you're bound to miss the gorilla.
Biology is the king of this type of science: you're studying a chaotic system with more interactions than you can imagine. Just pointing out that something unexpected is happening is perfectly publishable.
I'm sure you can staple on a null hypothesis after the fact - "Our null hypothesis is 'woozits never interact with greebles' and oh would you just look at that" - but that's always rung hollow to me.
Also, methods papers. Scientists are basically just advertising their tools to other scientists. Again, I suppose you could ram in a null hypothesis of "it doesn't work any better," but while methods papers do need to falsify that, it's not really the established aim.
This isn't applicable to the ufo situation because it's not a novel phenomenon. We can already study UFOs, and every one we've gotten a good look at has turned out to have a mundane origin. No reason to suppose the others aren't mundane as well.