criticisms of that poll.
I think it does, however, also show that the margin of error in perhaps all other polls that use weighting and algorithms to make up for too small a sample size is probably significantly larger than they indicate.
I've seen polls released over the same time frame that had very large differences from each other, but both claimed a much smaller margin for error than that discrepancy.
North Carolina
http://elections.huffingtonpost.com/pollster/2016-north-carolina-president-trump-vs-clinton
Trump +7
SurveyUSA/WRAL
Oct 28 – Oct 31
659 Likely Voters
Clinton +7
SurveyMonkey
Oct 26 – Nov 1
1,617 Likely Voters
FLORIDA
http://elections.huffingtonpost.com/pollster/2016-florida-presidential-general-election-trump-vs-clinton
Clinton +8
TargetSmart/William & Mary
Oct 25 – Oct 30
718 Likely Voters
Trump +4
Remington Research Group (R)/Axiom Strategies
Oct 30 – Oct 30
989 Likely Voters