I believe there's a math error somewhere in multiminer.  I downloaded Nemos (which looks like it was written by the same authors as Multiminer) and it's benchmarking right now.  So far, the btc/day numbers are just about half that of Multiminer and I'd say about 95% accuracy to what I've been earning on MPH over the past 4 days.  I posted my last log file on Git as you requested.