there is also an interaction between D (source life) and E (stability) ((and the number of mats and time remaining- but that seems to be a second order effect))
For instance in mode 3, the more that stability is greater than life (E>D) the more effect there is (on average) on life vs stability.
In mode 5, as long as E>D both E and D tend to be affected together. Once E < D E is effected much more by each action.
It's an interesting hypothesis, but not one borne out by the RyzomCore source. In all the relevant code paths, D and E are independent and do not couple to each other.
Some of your observations may have alternative explanations that don't contradict the available information. Take mode 5 for instance: D couples to speed, which will end up somewhat more strongly damped (21 coefficient and 4x absorption from matspec) than E, which is coupled to quality (39!! coefficient and 3x absorption from ecospec). But with all the absorption being slung around the impact ranges may end up similar enough to each other that you'll frequently observe the bars going down together at first, until E takes (and statistically almost certainly keeps) the natural lead.