It wasn't so much that engineers didn't know how to master, but that the analog-to-digital sampling was done at the CD bit rate (44.1Khz). That meant that the analogue signal needed to be brick-wall filtered to remove any information above 20KHz, and analogue brick-wall (hi-Q) filtering does really, really bad things to time and phase. Since the late '80s, the sampling rate on conversion has been much higher and the removal of high-frequency information (possible aliasing) is done in the digital domain. (And it really didn't help that early players did no dithering on playback either. Early CDs still sound horrible, but you really need to play them on early players to get the full dentistry-without-anaesthetics effect.)
Yes. That's the reason why I put converters first, engineers second. I shouldn't have said engineers though. What I really meant was the peripheral equipment they didn't have at hand to provide a proper transfer to the digital domain.
If I'm not mistaken, when sampling @ 44.1k your analog filter has to go from 0db @ 20kHz to "nothing" (-20dB or even less) @ 22.05k. Of course it doesn't (or at least, not nicely)
Now, if you record at 96k your analog filter has to do that from 20kHz to 48kHz (and that's not really needed most of the time as there's almost nothing there to begin with)
Downsampling digitally from 96k to 44.1k is easier and with much more control over the process.
