I am not sure if I understand the graph, but I assume the y-axis is capacity and the x-axis is time (it would help to have a scale :-)
Anyhow, just wanted to report that after ~12 dives with the modified flash I have many TTL shots that came perfect on and some that did not. I am not sure if this is due to my errors (e.g., positioning of the strobes) or due to the flash output capacity. I did not hold a careful comparison.
Here are two examples, shot with TTL with the 10uf mod and two external strobes (Sea & Sea YS110a + DS-Y1, both set on TTL). Didn't change the exposure in development from RAW
In some cases I had to set the exposure higher in Lightroom. So the mod might be under-exposing in some situation.
Regarding my earlier comparison, I did repeat a similar test with a full housing, and it seems that having the fiber-optic cable in the right positioning helps... and the results are better.