I may be reading the poll wrong but it seems this poll has it Trump +2
No it's not. Results are not uniformly distributed within the margin of error, they are normally distributed. That means that Monmouth is estimating Trump +2 as the most likely outcome, not a tie. There's a certain chance that the sample is drawing from a 'tied' or better population, but that is less than a 50% chance and also the same chance the sample is drawing from a Trump +4 or better population. Consequently, Monmouth going from tied to Trump +2 is not them getting an equivalent result. It is them estimating a 2 point swing to Trump. Again, there's a certain chance that the population has not changed, but a much greater chance that it has.
