I am very sorry to hear that Mr. Travers is no longer with us. I only know of him through his forum posts, but if they are any indication those closer to him have lost quite a person. My condolences.
It appears from the experiment results that my initial theory must have been incorrect, but I'll state it here for the record anyway. Later in the post I'll take a stab at modifying the theory based on the evidence:
Original theory: The amount of pain we feel from a joint lock is inversely proportional to how much resistance we can offer. In other words, if we are standing on solid pavement we should be able to exert more force through our frame to counter the locking action-- unless our partner gets effective kuzushi (balance breaking) at which point we can offer little or no resistance and therefore feel the lock more strongly. The theory was that if the lockee is on the grass he cannot resist the lock as effectively because the ground absorbs some of his energy and also because the softer surface causes him to lose his balance sooner. Based on this idea, the most painful combination should have been lockee on grass and locker on pavement. It appears that was not the case at all.
New theory: 1. People tend to do things with their body when they apply joint locks that make them less effective. 2. Joint pain results from having all the "slack" taken out, then having the lock applied still further.
Softer ground would tend to minimize any errors (such as tensing the legs and lifting the body rather than settling) that the person applying the lock might make. In other words, it sort of forces that total relaxation we strive for, and forgives us if we have more tension than we should in some parts of our body. On the lockee side, being on pavement affords no additional slack once the lock is applied. Softer ground might give a bit and absorb some of the power applied, but no such luck on hard surfaces.
So that's my "non-ki" explanation attempt for why the worst case is locker on grass lockee on pavement. It's only a guess, but I think all this stuff is really interesting (joint locking is one of the neatest aspects of aikijujutsu, IMHO, and I am studying it very closely these days). I'd love to know what the rest of you think about how or why this works, including any ki explanations.
PS: Sharon, thanks for conducting the experiment for me-- I don't know anyone who will let me play with joint locks outside of class.
PPS: Do you think the experiment results would change if repeated with resisting and non-resisting lockees?