The 54-45 numbers indicate that the poll was one that forced respondents to choose one of the candidates. One reason for doing this is to try to capture which way the “not sure” voters are leaning. I don’t know enough about polling to know if the forcing actually improves accuracy this far out. I could easily see it pushing respondents toward a more-familiar candidate early on.
In general, though, I think it is more helpful to see how the numbers from an individual pollster are trending over time than it is to compare numbers from one pollster to numbers from another pollster. If you see polls from most pollsters trending in the same direction, you can have some confidence that the trend exists.
Agree with all you said, especially including the conclusion. No matter how good Beto is or how much enthusiasm TX Democrats have for voting this year, this was always going to be an uphill battle that Beto could win only with a huge amount of work and a large dose of good luck.