Sure yeah, if you can find sources that backup being able to improve mitochondrial efficiency, I think it's a good idea.
For calculating the diminishing returns, we could use the formula we planned for agents:
Eff = l - (l-s) * a^(-n)
l - Limit. The point it ends at, in case we decide to lower that value.
s - Starting point. What f is.
a - How fast does it grow. Higher values make it grow faster.
I would say for mitochondria, chloroplasts, and thermoplasts, the upgradeable stats are efficiency and rate.
For flagella and cilia, the upgradeable stats are speed, turn rate, and efficiency/energy cost.
The cell membrane, cell walls, pseudopodia, and lamellipodia seem like a big topic so maybe save that for later.
That leaves the nuclear core, vacuoles, cytoplasm, predatory pilli, and bioluminescent organelles.
NOTE: Also, lysosomes have not been discussed much. Are we assuming they're there? Will we include them?
Just to demonstrate, if we used the maximum you provided, as well as say a minimum of 20, for the ATP to be able to be produced, and we set a growth constant of 1.3, the results would be:
Also I was thinking, for upgrades to organelles, I think they should apply across the cell as a whole, so instead of having to upgrade the efficiency of each mitochondrion, it would just upgrade them all at once.