That was a misleading statement on the part of the person who wrote the article.
If you go to the article about the Virginia race, it says the interviews were with REGISTERED VOTERS who said they were likely to vote in November. I don't believe that is the standard way to qualify someone as a LIKELY VOTER. It takes more than simply asking a person if they are likely to vote. You have to inquire about their behavior. Gallup uses 7 questions as I recall.
If you look on the Real Clear Politics website, the Mason Dixon Virginia poll is classified as a Registered Voter (RV) poll, I assume for this very reason.
Oh, ok -- thanks for telling me that.