I can't speak for the Apple system but you CAN have lossless compression. It depends on the algorithm used. However, truly lossless compression generally can't make files as small as lossy compression.
For example, consider a 16 bit audio file--a string of 16 0s and 1s. By the law of averages, a certain number of samples will begin with a long string of 0s which contribute nothing to the sample. If you can use variable length coding and discard any initial 0s, you can compress the file size but not lose any significant information.
Lossless coding can go farther than that. If you have a series of identical samples, there are ways to send the information once with an electronic "repeat for X samples" command. Again, this can reduce file sizes without making any un-reversible changes to the sounds. For example, if you had a 1000Hz tone at a specific level, you could (in theory) send a single sample and reproduce a hour of that tone from the one sample.
Obviously, real music doesn't work this way and the amount of compression available with lossless techniques can be quite variable depending on the audio content--but things like FLAC can genuinely be lossless.
However, for really significant compression, things like MP3 start playing lots of psychoacoustic tricks--and throwing away lots of audio data. It also uses all the "look ahead" and predictive coding tricks you mention--basically it guesses what the music should be and, like any guess, can get it wrong! This means that the effect of lossy compression is cumulative and it starts to bite seriously after a few passes. Frankly, I only use MP3 when I need to email a file or something at the end of the recording and mixing process. It certainly has no place in the production process.
