The problem with the poll is the equal allocation of votes counted to each state. Each state gets 1000 votes in the poll. The state by state polling could well be accurate. But that cannot be translated to a national poll by simply adding up the votes.
There may well be more Republican states than Dem states in which case the poll results would predictably favor Trump. I suspect that is the case here. We hame more states. The Dems have the big ones (mostly.)
You’re right — votes/state does not translate into Electoral College votes. However, it’s not at all unusual to start with the same sample size/state — in order to get to the same confidence interval. Results can then be weighted by whatever factors one chooses.
There are all sorts of other considerations: breakdown by sex/race/ethnicity/sexual orientation/income class/etc. Those considerations can affect sample size. If you want (e.g.) to compare male to female voters, you will have a larger margin of error for each subgroup — unless you increase the sample size. When you adjust the sample size for any subgroup, you have to use weighting factors in analysis, to avoid distorting your results.
It’s really impossible to say whether or not the survey in question was done right; without seeing their complete methodology. Just from the article, I can think of a several possible threats to the validity of the results (you’ve mentioned some of them)— but we just don’t have enough information to be sure.