Standard audio CDs are defined by the "Red Book" standard. Red Book specifies no filesystem.
An audio CD is divided into tracks, then subdivided into frames. Each frame represents 1/75th of a second of audio. Each frame contain 2,352 bytes of data. All 2.352 bytes are devoted to audio, no positioning data is included (i.e. no filesystem). This isn't a problem for an audio CD player - it just reads a track straight through from start to finish. It has no need to jump to precise locations (+/- one bit) on the disc.
Data CDs (ISO-9660 standard) include a filesystem. Frames are still the same total size, 2,352 bytes, but only 2,048 bytes are data. The remaining 304 bytes are indexing information that allows the disc reader to skip to an exact location on the disc. This is necessary because computers must read data in chunks. (The data busses are shared amonsst the various devices - each device must take turns, using the bus for a brief period). The positioning info interspersed throughout the data makes this possible.
Rading an audio CD (which lacks positioning data) on a machine that must frequently stop and restart the read operation is problematic. Wtihout the position idexes mised in, the best a CD-ROM drive can do is guess at where it last read from the disc. Often times it guesses wrong. Bits are misread, errors are introduced, and the audio quality suffers.
These read errors cause sharp transients (clicks and pops and such). If two non-sequential bits are read as sequential, the audio signal will jump instantly from one level to another without passing through any intermediary levels.
Most software has a crude way of dealing with these errors - it simply filters out any sharp transients. This eliminate the clicks and pops that the average users find annoying. But it also gets rid of the fine details that the critical listeners expect to hear.
Quality software can make a world of difference. A good piece of software has the ability to overcome these limitations effectively. But these algorithms aren't simple, and they don't come cheap. Most of the algorithms to do this are copyrighted, and aren't included in any of the free mp3 encoders I've seen.
Any way, the bottom line is that there's a whole lot of technical difficulty involved in getting quality audio off of a CD and into a computer. Music from a PC is pretty easy to screw up (in an audiophile sense - most typical users would never know the difference). Condemning .mp3 because one particular listening test sucked doesn't really make sense.
But if you get the hardware and the software right, .mp3 audio can be higher quality than CD audio.
(Yeah, I'm a real geek.
)