First, you check for external factors which may have affect the test. If you can confirm you cannot reproduce past results, you may consider discussing with other people about your findings, and publishing if no one can find a reasonable explanation. In any case, you are happy because that's how science progresses.
> My position is that I am happy to continue as before, but I won't overstate my case. I won't say 'I know' when I don't. I can accept that what is presented as true, may in fact be true, but I don't know it. I would be lying if I did.
Sure, and that's the point of science: there is no truth, only better and better models. But we won't stop using the verb "to know" altogether. Knowledge is not a binary thing (you know or you don't).