One way to deal with this is discussed in the article -- if details of reports are not widely publicized (often the case), then details of reports/allegations can be compared. In many cases, as noted in the post, methods are consistent. You could use some Bayesian analysis to figure out the probability that independent reports would come up with the same details.