Actually it's arguable that the MFSL Steely Dan waveform is not ideal either as it leaves some dynamic range on the table by being "too quiet". Those max peaks come nowhere near the max of what they might. You can bring up the maximum peak levels almost to the absolute maximum before clipping and not freak out most modern DACs, but even you took it down to say 95% of max you'd get closer to full dynamic range.
To be clear, I'm not talking about adding any compression, just making sure you use all the available dynamic range on the CD. That MFSL is fairly old, when the Sony spec said not to raise levels above a certain point so you don't flip out early DACs. Steve Hoffman figured out early on that warning was bunk; if you look at many of his mid-80s masterings, they peak out in the high 90s or even 100%.
Actually, there is a disadvantage to peak-normalization(applying gain so that all songs
on an album peak at 0dB full scale(or .5dB below that): Humans do not judge
loudness by peak levels; they judge it by average levels(close to but not exactly
RMS). So with tracks peak-normalized on the CD, some tracks will sound louder than
others, requiring frequent volume adjustment during a whole album listening .
Additionally, peak normalizing to the least dynamic track, you would need to
peak-limit a more dynamic track, throwing away all those peaks & dynamic range,
just to make it sound as loud as the more compressed one.
When mastering, I would adjust the track levels on the album by how loud
they sounded to me, not by how high they peak. This means some less dynamic
songs may peak highest at -6dB FS, yet *sound as loud* as other, more dynamic
tracks on the album which peak closer to full-scale. This ensures less need for
the consumer to adjust their volume after every other song, especially while
concentrating on a task such as driving.
Look up Bob Katz's "K metering" system for the plug-in that assists with mastering
levels set by loudness and not by peak.