.
 
 

How are you calculating those numbers?
 
The formula used in audio to convert from dB to ratio is ratio = 10 ^ (db's/20)
 
Perhaps you were using a formula from somewhere that divided by 10 instead of 20 for most of those calculations?
 
Confusingly, decibels can be used to describe different things, and if you're dividing the #dB's by 10 instead of 20 (or multiplying the log10 of the ratio by 10 instead of 20 if you're converting ratio to dB's), that's for decibels expressing a power quantity which is not what we use for audio - we use root power or field, in which case we use the formulas that divide or multiply by 20. See here:
 
https://en.wikipedia.org/wiki/Decibel#Definition
 
It's confusing, but you really just need to remember that for audio we use the formulas with a "20" and not a "10" if you see dB/ratio conversion formulas somewhere.
 
So -6dB = 10 ^ (-6/20) = 10^(-.3)=.5012
 
and -20dB = 10 ^ (-20/20) = 10^(-1)=.1
 
But for some reason, your figure for 16dB is essentially right. Did you use a different formula there?

You do need a perfect band-limited filter if you want a perfectly reconstructed signal. I'm not debating whether one is needed in audio or not, just stating the obvious that the sampling theorem does not provide a simple, straight forward way to determine the correct minimum sample rate for a system without a perfectly band-limited filter. The interpolation I was referring to is in reconstructing the signal in the DAC from digital data to analog signal. Interpolating between samples. I was speculating that this might be playing a role in the up-sampling soft synths that Craig and others are discussing. That the physical reconstruction filter itself is playing a role in this phenomena not necessarily so much the up-sampling. That a better designed interpolation filter might provide the same auditory stimulation in humans.

Its not necessarily what you hear verses the accuracy and potential artifacts etc from processing.   Higher bit rates  apparently allow for more transparent processing so when the mastering or mixing process is complete the sound of the tracks are not adversely effected by all the manipulation by plugins and other forms of digital modifications  At the end 44.1-16 is more than enough for listening 

1's...0's..X's...O's...
Just use your ears, that's what everyone else is using to listen to your music.

chuckebaby
1's...0's..X's...O's...
Just use your ears, that's what everyone else is using to listen to your music.

Actually they're using their own ears. Not yours.
 
That's why the word "objective" comes into play for some of us.