ANY poll testing the national popular vote in close elections is going to be off because winning or losing the popular vote means nothing constitutionally.
Only a poll of all the separate states can gauge close elections, IMHO.
State polls were all over the board. On the same day a couple weeks before the election for example, 6 NC polls came out within 24 hours - 2 had Trump up big (4-6%), 3 had Clinton up big (5-7%), and 1 had it tied. If you think HRC's popular vote lead is +10%, then HRC up 5-7% is likely right. If you think the race is close to tied, it's safe to throw out those 3 state polls from NC that had her up massively.