First of all, there is no lossless setting on MP3s. Even pegging the bitrate at max (320 kbps) won't accomplish that. Second, the CD standard has been criticized by audiophiles since its inception. The sampling rate is not nearly high enough for representing the highest audible frequencies. The best it can do theoretically is a sawtooth wave at 22 kHz, and the distortion is still very high at 15-18 kHz. The brickwall filtering required to eliminate all the ugly artifacts itself introduces other distortions, such as phase anomalies. A much-higher sampling rate is needed to shift such problem well above 20 kHz. 96 kHz will do much better than 44.1 kHz, shifting most of the problems well into the ultrasonic. If we can afford 192 kHz, why not?
The format's other problem is the sample size. 16 bits does not do the job for anything with a realistic dynamic range. Compressed popular music is served OK by it. But classical music and uncompressed music of any kind will suffer in a way which is very unnatural. In effect, the distortion curve is upside-down. In a digital recording, the softer the sounds, the more distorted they become. That's because as the amplitudes approach silence, there are much less digital steps to approximate the waveform. And in order to fit these very soft passages along with the bombastic ones, without sacrificing the real dynamic range, there is no choice but to adjust the recording level so that the loudest sounds don't clip, and make do with the jaggy distortion at low levels. This can be reduced by adding a bit of white noise. Ironic, no? The format originally touted for its lack of hiss has to have some hiss added to better modulate low-amplitude waveforms. Going to 24 bits would increase the amount of quantization levels 256-fold, which would be more than adequate to sink low-amplitude distortion into inaudibility.
