I've found this paper enlightening, as it gives a wide, though not complete, benchmark for different sequence map algorithms. As we are dealing with a number of expression profiling experiments, this paper will be surely helpful in choosing alternatives to ELAND.
Besides, it would be interesting to compare ELAND itself with the algorithms discussed here. I suspect that the well known "2 mismatch limit", in addition to the lack of gapped alignment capabilities, may make it perform not better than the worst cases discussed.