I just checked with version 1.5. The ‘opt max throughput’ scheduler looks fine to me as it does not assign all resources only to one user. I think the effect you are referring to ist, that the curve for the user throughput shows zero throughput for all users except the first one at high SNR. This happens due to the implementation of the exhausitve search algorithm over all schedules. Once, a schedule with maximum throughput is found, and there is no other better schedule, it is taken as user assignment for the next transmission. At high SNR however, all users will be able to user CQI 15 in all resource blocks, meaning the maximum throughput is obtained if any user is assigned all RBs. The first schedule considered within the exhaustive search algorithm that does this, is the one where user 1 is assigned all resources.
This is our course not fair at all, but yields maximum throughput which is the ultimate goal of this schedulung algorithm. This scheduler does not consider fairnes at all. To improve fairnes at high SNR, one can add randomization for these high SNR cases.