Wikipedia:Verifiability, not truth stresses verifiability as a necessary condition for inclusion. As the essay also notes, verifiability is not a sufficient condition for inclusion. Indeed, several Wikipedia policies address when verifiable materials should be excluded. One might say that at least some of these policies rely on original research ("OR")—we don't, for example, require a reliable source to call a theory "fringe" before WP:FRINGE applies, and rightly so. But original research is not always an appropriate basis for exclusion. Specifically, original research concerning source accuracy should be discarded in almost all circumstances.
Source-accuracy considerations are distinct in kind from other rationales for exclusion. They require a level of expertise that goes well beyond surveying the available sources. To illustrate, consider how mediation of source-accuracy dispute would play out: While an RFC on whether a viewpoint is fringe would be fairly straightforward, imagine an RFC on the merits of a medical journal article's methodology ... or one in which the community is asked to settle a debate among two rocket-scientist Wikipedians. Would participants in such a discussion have to show their work? Non-STEM subjects might seem like safer ground, and it's on these subjects that accuracy-related OR is most often argued as a basis for exclusion. These arguments can take several forms—usually, they're an allegation that reliable source was either taken in by a liar or otherwise failed to perform due diligence. Sometimes, these users might be right, and they might be able to make a particularly convincing case. But this is, in effect, armchair quarterbacking. The very fact that non-STEM subjects can seem more "doable" to non-experts is, itself, the problem, as non-experts will be more confident about their ability to analyze a non-STEM issue regardless of whether that confidence is correlated with competence. And the difficulty in identifying when expertise is required cautions against a case-by-case approach, which is precisely why we don't consider the strength of an argument in applying our bar on original research as a basis for inclusion.
Our job, generally, is to reflect what's stated in reliable sources (with only certain exceptions). We abide by that responsibility because the accuracy of reliable sources is thought to be superior to the accuracy of an individual Wikipedia editor or even a group of Wikipedia editors. Fortunately, there are usually other reasons the exclude false info (see, e.g., WP:FRINGE, WP:MINORASPECT, or WP:NOTEVERYTHING). But, absent the applicability of another policy, attribution to a reliable source should be a sufficient basis for inclusion; we should not exclude on the basis of our personal determination of accuracy. No doubt, adhering to that philosophy can be frustrating, especially if you know the sources are wrong. But this project, generally, has faith that reliable sources (or, at least, some reliable sources) will, with time, get an issue right. By following "verifiability, not accuracy", we can better ensure that we'll eventually get it right, too.