For a bit, I thought that this seemed to set a floor on the number of respondents who must really be producing memories of the fake event (although significantly more people who claimed to remember the fake event picked it out as fake than would be predicted by chance if they had no way of distinguishing between them,) but it occurs to me that if the respondents were lying about remembering multiple events in the survey, then it doesn't seem that the survey would be able to distinguish between respondents who picked the wrong event as fake because they had produced memories of it as vivid as of the others, and respondents who picked the wrong event as fake because there were multiple events on the survey that they didn't remember, and their attempts to determine which one didn't happen come down to luck.