Looking at the # of people who responded, I see that it was less than 300 people.
That's a pretty small sample.
I don't know how it aggregates with polls that include a wider sample of respondents, and I don't know whether they were calling landlines or land and cell phones.