The pertinent bit from the actual poll:

When asked about their preferences regarding the upcoming presidential election, 46.3% of respondents would vote for Donald Trump, and 36.5% would vote for Kamala Harris. Another 6.8% chose neither candidate, 7.3% did not plan to vote or cannot vote, and 3.1% did not know (Figure 11). After excluding those who do not plan on voting or cannot vote, 49.9% of respondents would vote for Donald Trump, and 39.4% would vote for Kamala Harris. Among registered voters after excluding those who do not plan on voting or cannot vote, the support for Donald Trump was slightly lower, with 48.2% of respondents supporting him and 43.2% supporting Kamala Harris (Figure 12).

So the five percent margin assumes none of the “don’t plan to vote” respondents vote anyway, that none of the “not registered” registered after getting polled, that the undecideds don’t break for Trump, or that any of the “can’t vote” respondents actually can. This also wasn’t a traditional presidential poll but rather a general statewide opinion poll done by a university so make of that what you will.