My guess is that a lot of the psychoacoustic models used by lossy codecs aren't tuned for electronic music. The resonant distortion on the bassline of the linked track, for example, might sound different in the original. I have one song in which I can tell a difference for sure: Who Said by Planet Funk, about 1:30 in, has a rising modulated treble sound that sounds awful in 320kbit MP3 compared to the original.
Psychoacoustic models are very good, but MP3 has unavoidable coding weaknesses. Use a modern audio codec and you won't have this problem nearly as much.
