It’s actually mentioned in the article:
“There are a couple of goofy outliers (one person labeled “probably not” as 100%) but zonination says the outliers are all from the same respondent (so they probably should have been omitted).”