I agree that lukeprog's examples don't really support his point very well. However there are other examples that do so better, that do not fit your criterion of "if we consciously think about them, we realize our mistake" - cases where people had to be actively convinced through experiment of what was really going on, denying it for quite some time. For instance, to reuse the same quote from Schwitzgebel I posted in another comment:

You might think that the blind, whose abilities at echolocation are generally thought to be superior to those of normally sighted people, and who often actively use echolocation to dodge objects in novel environments, would be immune to such ignorance. Not so. For example, one of the two blind participants in Supa and colleagues’ 1944 study believed that his ability to avoid collisions was supported by cutaneous sensations in his forehead and that sound was irrelevant and distracted him (p. 144 and 146). Although asked to attend carefully to what allowed him to avoid colliding with silent obstacles, it was only after a long series of experiments, with and without auditory information, and several resultant collisions, that he was finally convinced. Similarly Philip Worchel and Karl Dallenbach (1947) report a nearly blind participant convinced that he detected the presence of objects by feeling pressure on his face. Like Supa’s subject, he was disabused of this idea only after long experimentation. (This participant, it turned out, used his impoverished visual sense of light and dark more than tactile or echoic information.) Such opinions used to be so common among the blind – until Supa, Dallenbach, and their collaborators demonstrated otherwise – that the blind’s ability to avoid objects in novel and changing environments was widely regarded as a tactile or tactile-like “facial vision”, perhaps underwritten by feeling air currents or the like (see Diderot 1749/1916; James 1890/1981; Hayes 1935; Supa et al. 1944; the negligible relevance of air currents is shown by participants’ excellent performance when ears are uncovered and cloth is draped over the rest of the face and their poor performance when ears are stopped and the face is left clear). Presumably, if blind people experience auditory echoic phenomenology, and if they are – as people in general are widely assumed to be – accurate judges of their phenomenology, it should occur to them that they detect silent objects at least in part through audition. They should not make such large mistakes about the informational underpinnings of their object sense.