The primary error there is one that people can't seem to stop making for some strange reason: if you don't measure a property X, and instead measure a (non-commuting) property Y, you are not free to say what X "would have been," and -- especially egregiously -- to use that to demonstrate a contradiction. There is no "would have been" in QM. Either you measure it or you don't, and even after you do, the value becomes invalid after it evolves or a complementary variable is measured.
Secondly, Wigner's friend was not a single particle. Asking what a single particle "experienced" only makes sense if you are a panpsychist (or "panpsychic" as amusingly misspoken in the article). In any case, it has no bearing on the essence of Wigner's question.