Not related to these specific numbers, but what is thought to be Gravis' main "flaw"? It's constantly made fun of as the worst polling service. Do they use a bad method/generalize their questions/use bad sampling?
They had a horrible track record in 2012. It was slightly better in 2014, but their polls seem very amateurish, misspelling the names of candidates, using obviously absurd demographics (such as the NH D primary being 56-44 male when it was 58-42 female in 2008), and even their own polls being inconsistent with each other. They've shown Bush doing better in SC than in FL, Warren doing better in NH than MA, showing Rand Paul beating Warren by a bigger margin in CA than in NC, etc. etc. They also "recalled" a poll in 2014 that was an obvious outlier. Overall, they're extremely shoddy.