You'd have to ask Knuth why he thinks so. He may have seen the source code, for example.
Speaking merely as a user who remembers when the Mac was first introduced: Because it was a program which you could play with for sixty seconds and immediately get a vision -- not only of exactly what the Mac was all about, but of what the next few decades in computing were going to look like. It's hard to describe the intensity of the experience of seeing it for the first time. I still have the issue of Popular Science in which I first saw a photo of the interface.
Sure, in a sense it was nothing that Engelbart and the PARC guys hadn't done before. But it was no research project; it was a shipping consumer product with a solid, elegant, friendly, minimalist design, built (according to Wikipedia) in 8500 lines of source code, that ran crisply on a PC which cost far less than the $10k Apple Lisa, to say nothing of a $100k piece of custom Xerox research hardware.
With all these fragmentary, unsourced anecdotes about Donald Knuth, one would think he hadn't given numerous interviews and written several actual books.
