This is a poll of randomly-selected adults. They're not even necessarily registered voters and probably the pool of likely voters in this bunch is less than half of those interviewed.
I'd love to see his approval ratings in a poll of likely voters. Gotta be less than 40%.
Important point!
Yes, you make excellent points. This is a major flaw in Gallup's traditional methodology for their political polls: they don't sample likely voters only until at most a month or two before the relevant upcoming election.
In fact, I'd go one step further. They don't even screen for American citizenship in their poll respondents; their sample group is adults 18 and over living in the United States. No doubt that the flaws of their technique gives Democrats artificially inflated numbers compared to what they would get in a poll of likely voters.