Hi Fumbles,
Sorry for the VERY late reply. I didn't have internet access for a long time. The short answer is YES, most HDCD's are really closer to 16 bits than 20. Here's the long answer:
Only two of the encoding features affect the dynamic range -- Peak Extend and Low-Level Extension. If Peak Extend was turned on by the mastering engineer, the top 9 dB of level on the master recording is "scrunched" into only 3 dB of level on the disc. Since most pop and rock is ALREADY over-compressed, the instruction manual for the HDCD encoder recommends NOT to use Peak Extend for material that is already heavily compressed. If it is, when played back on a non-HDCD player, it will have almost NO dynamic range and sound pretty bad.
But if a recording uses the Peak Extend feature, it can (at the most) result in a 6 dB increase in dynamic range when played back on a player that decodes it. Now we have to do some math. Each doubling of signal level is equal to 6 dB, which is the same as 1 bit. So if used properly, the Peak Extend can increase the dynamic range of a CD from 16 bits to 17 bits when decoded.
The other feature that affects the dynamic range is Low-Level Extension (LLE). As the recorded signal drops below -45 dB below Full Scale (dBFS), the gain or the recording is increased in 0.5 dB steps. But the MOST gain change that can possibly happen is only 4.0 dB. This is not even 1 bit's worth of dynamic range. So the VERY MOST increase that can be had is actually only about 1.5 dB, NOT the 4 dB that HDCD claimed. (Remember when I said they had great engineers, but sleazy marketers?)
But even more discouraging is that LLE is virtually useless. It only comes into effect at such extremely low levels that almost the only time that it is used is in the fade out between tracks. I have only seen the LLE light come on an a VERY few classical recordings of solo guitar or harp or harpsichord or things like that where the levels are EXTREMELY low. And even if it was used, the most increase it would give would be around 0.5 bits!!
So you are probably wondering why in the world they claimed up to 20 dB of dynamic range. Well, politely is is called "smoke and mirrors". Bluntly it is called marketing BS. They have six or seven selectable dither curves when recording with the HDCD converter. Dither changes what would be distortion that is correlated with the music signal into just random noise (like tape "hiss"). They claimed that this would yield another two bits of effective dynamic range. Well, this is pretty debatable and it depends on how you define dynamic range. But the kicker is that nowadays virtually every A/D converter in the world has dither. And even before they intentionally added dither, the early A/D converters had enough noise in their analog electronics to produce what actually turns out to be a very good dither signal!!
The bottom line is that if you have some way to measure the dynamic range, you are almost NEVER going to see anything going past 17 dB. That is because the only process that reliably adds any dynamic range is the Peak Extend, and it only adds 1 bit's worth of extra range.
I hope this clears up what you have been seeing with your analyses, which are very much correct -- and far more accurate than the claims made by the HDCD marketing department!
The best thing about the HDCD encoded discs was that for about 10 years, it really was the best sounding A/D converter on the planet. From around 1995 their one-of-a-kind prototype machine was built it took almost ten years before the other manufacturers of pro gear caught up with the sheer sound quality of the HDCD converter. Keith Johnson really went all-out on that thing and it was truly amazing in its day -- but the great sound quality had virtually ZERO to do with the HDCD features....