I think at least some of the answer may be further down in that post from Straight Dope that babalu87 was pointing at:
Apart from incombustible elements, the color of a hydrocarbon flame is primarily dependent on the richness of the flame – that is, on how much oxygen there is to combust the fuel. In practice*, when the mixture is slightly lean (has more oxygen than required for complete combustion), the color of the combustion zone is generally blue-violet due to large amounts of high-energy radical carbon and hydrogen compounds. When the mixture is slightly rich (slightly too much fuel and not enough oxygen), the color is sometimes green due to C2 molecules breaking free, and the high-temperature products can glow red from the CO2 and H2O produced during combustion. When the mixture is very fuel rich (a poor flame, with not enough oxygen to burn properly), carbon particles form and an intense yellow radiation results from their being heated in the flame. In very rich flames – often you see this in candles – soot particles may impart a black color to the outer edge of the yellow flame.
This would suggest to me that the logs putting out yellow flames are outgassing more, giving a richer mixture than the logs putting out less gas at any given moment (assuming the air inlet is the same setting)
This would be consistent with the notion that softwoods burn hotter but faster (more gas for a shorter time) and put out more creosote (less complete combustion due to insufficient air in the mix) Once the initial gas is burned off and you are down to coals, then you will see blue flames because you now have plenty of oxygen, with the flame size being limited by the amount of fuel gas coming off the coals....
This is just theory, but it seems to fit to a first approximation.
Gooserider