Some polls near an election predict the vote. Some polls not near an election identify the perception of the president.
My point is that the methodology used by the pollsters for sampling seem to be pretty accurate, since the election polls (consensus) was close to the outcome (within a percentage point). Thus, the approval polls are probably pretty accurate. As far as whether they are "likely to vote"- it doesn't matter. The "approval" polls are not really predicting an election. It doesn't matter whether you are likely to vote. If the people hate your leadership, then they hate you, whether they vote or not.