The link you want is here:
http://irc.nrc-cnrc.gc.ca/pubs/ir/ir766/
The data will also show(top of actual page: 88) that there is a point difference in STC and/or IIC depending on placement of absorption material.

If points matter...
The data becomes clearer as you perform many tests. Comparing one isolated test vs. another may simply be comparing anomolies. I look at the quote from AA Warnock:
Position of sound absorbing material in the floor cavity:
To test the effect of moving the sound absorbing material inside the floor cavity, 152 mm thick glass fiber batts were placed at the bottom (against the ceiling), in the middle and at the top (against the subfloor) of the 235 mm deep cavity of the reference floor. As expected, changing the position did not change the results.
Thing is, this is a good document, maybe not for this specific topic as it was written based on data produced by many different assemblies of many different materials in many configurations. So a broad statement taken out of context cannot be made with this document as support.
Page 25 will disagree with you. A thickness of 90mm to 456mm(3 - 1/2" to 18") shows an increase in overall performance in both STC and IIC.
No disagreement. I wasn't comparing R13 in the floor. What I stated was that there was little difference between an R19 filled cavity (152mm) and a full cavity, in this case 456mm. The majority of the work is done with 6" of fiberglass. This is especially true when we look at a heavier system where the added mass would alleviate more of that STC difference. Lastly, you're looking at I-Joist data which they identified as being wildly anomolous. Better to look at the data on page 17, which is solid wood joists which are far more consistent.
It also will prove my statements incorrect about "bridging" from over stuffing! Whadaya do? No, you're right. Overstufing will most certainly couple the surfaces. More of a risk when you start with demser insulation.
I appreciate what your doing. I love this stuff too! Thanks for the comments. I always pick up something new when looking at data again
Brien