To answer one of my own questions, if you alter the percentages - which you have the freedom to do - away from 1/3, 1/3, 1/3 to, say, deciders: 5% compensators: 90% benchmarkers (which I'm confused about): 5% then the compensators will have a fixed fee per capita which is small (about 18 times smaller than average). Therefore, my complaint that "poor" voters would have incentive not to vote leaving to a rich-dominated electorate, will be diminished. For practical purposes in large elections this should be an adequate fix of this problem (if it really was a problem).
Designer has freedom to alter these 3 percentages to optimize real-world performance. ---- Election-Methods mailing list - see http://electorama.com/em for list info