When in science we measure we dramatically eliminate (as opposed to ordinary observation) our access to huge amount of potential information we can extract from the environment. Even so, as we have learned since the time of Huygens onward this is a price worth paying when our measures are theoretically fruitful (etc). One nice feature of measurement is that (if well designed) it makes observation of one's measurements trivial; so much so that one could, in principle, farm out the whole process of registering the measurements to machines (as is often done now). Observation does no interesting epistemic work in theory-mediated measurements. While there are sciences where observation matters a lot (taxonomy and Marc Hauser's experiments come to mind), for much of science (as in mathematics) it plays no foundational role whatsoever. Those of us working within General Philosophy of Science (GPOS) have known this since Duhem.
This is why Mohan and epistemologists more generally are wrong in thinking that Gettierized Epistemology (hereafter GE) provides the so-called content-general framework with which to evaluate knowledge claims and the status of science more generally. GE mistakenly presupposes that ordinary cognition, perception, and locution matter epistemically in science. Now this is not to deny that GE explores important issues that matter to us citizens on a day to day basis; the recently popular work on testimony comes to mind. (As long as the authority of science does not displace the experiences of ordinary witnesses this remains to be true in law courts.) But whatever special expertise GE has in answering the skeptic is largely irrelevant to thinking about science. Now (as I have remarked before) GPOS, by contrasts, rests on a quasi-transcendental assumption (QTA): if anything counts as knowledge it is fallible science, especially physics (chemistry, biology, whatever), so let's now articulate how this is possible or, more formally, justified. So, GPOS is a in bad position to answer the skeptic, except to remark that within the scientific machinery one is not much perturbed by it.
One crucial aspect of the Duhemian move is that the choice of measure allows one a degree of freedom that ordinary cognition does not. In particular, while coining concepts one can develop alternative or complementary theoretical structures with different measures--ideally their fruits will converge with or help refine our earlier measurements. Some folk working within GPOS think of this kind of (Wimsattian) robustness as an argument against skepticism, but this greediness is forgetful about QTA. Answering the skeptic is the domain of GE. In turn, GE oversteps it domain when it defends common sense against the authority of science (and often does so by pretending to speak the language of science).
A deeper challenge to GPOS comes from followers of Bergson (recall the discussion here, here, and here). Bergson agrees with Duhem that part of the legitimate success of our scientific edifices is the consequence of ignoring other possible ways of conceptualizing what Bergson calls multiplicity. Berson and Duhem allow that our present conceptual scientific scheme is dispensable for an alternative one. Not unlike Schlick, who argued that "the fact that [a system of science] is adopted by the whole of mankind and taught in all the universities would make no impression on me," Bergson (echoing Duhem) insisted that the consensus generating mechanisms of science shouldn't have the last word on truth. Unlike Schlick, Bergson did not rest his case on ordinary experience (and, hence, GE). Bergson argued for a special way of un-conceptualized knowing that is the consequence of scientific knowledge and the product of intellectual sympathy. That is to say, Bergson's approach can explain how the freedom that is presupposed in Duhemian GPOS is possible. While this is a freedom worth having, some other time I will explore if it can be actual today.