You said, “… my understanding was that convection is the most significant (and, in turn, it's driven by conduction and radiation).” It has been at least 33 years since your first premise was disproved both by calculation and experimentation. See:
Pelanne, C. M., "Heat Flow Principles in Thermal Insulations", J. of Thermal Insulation, Vol. 1, July, 1977, p. 49.
Tye, R. P., Editor, Heat Transmission Measurements in Thermal Insulation, ASTM Special Technical Publication 544, Chapter on Definitions and Thermal Modeling, January, 1980.
Both early studies showed theoretically and confirmed experimentally that natural convection within ANY fibrous insulation material is unimportant as compared to conduction and radiation if Ra is less than 40 for the material. Countless others have reaffirmed these earlier studies. The Ra number for fully lofted down averages ~ .18 and improves to ~.012 in the 2.5X density increase range. Natural convection loss has never been a significant loss factor even in the very first solid fiber synthetic polyester insulation bats. They had a Ra value of 6.79 which was very poor compared to down but good enough to prevent natural convection from being a significant factor.
Your second premise is also incorrect; “conduction and radiation drive natural convection." Neither are variables in the Ra related calculations. The topic is far too complicated to explain in this forum. If you want to understand the variables that factor into the Ra natural convection calculation, see http://en.wikipedia.org/wiki/Natural_convection
Your two erroneous premises did not lead you to a valid conclusion regarding the relationship between down density and thermal conductivity. I hope that you understand that I mean no disrespect, but I have grown weary from this topic discussion.