Why are high-profile misperceptions so persistent? Accurate information is widely available online and yet false and unsupported claims about issues like climate change and health care endure for years. This disjunction, we argue, is caused by ``correction mismatch'' --- people's lack of exposure to relevant correction information, which has been demonstrated to reduce misperceptions in past experiments. We examine this phenomenon in the context of false claims of widespread voter fraud with U.S. data from 2020--2022. Results using nationally representative experimental and observational data indicate that exposure to corrective information reduces fraud misperceptions. Turning to online behavior data we examine exposure to fraud claims at the article level and try to estimate to how often people encountered corrective information from fact-checkers or in mainstream news, especially among people who are predisposed to believe in these claims. The persistence of misperceptions, in other words, may be driven more by the information that people encounter than by their response when they encounter it.