There are no studies that have suggested anything like this. There simply has not been sufficient time to perform a study of the type needed to prove such a claim: you would need to gather a cohort of people that you know had the virus, wait for them to recover, then test for antibodies.

Moreover, even if the percentage of such people were shockingly high (i.e. over a fraction of a percentage), you would need to conduct a huge survey to distinguish it from error rates on the tests themselves.