How do they measure the number of cases that go either way? I assume this excludes settlements? If so it would explain the discrepancy. Arbitration is so much cheaper than going to trial that it would make sense to try a weaker case just based on risk vs reward.