ELAND is the Illumina GA Pipeline proprietary aligner for short read data. I can’t find much documentation on it, and I don’t have the code but this is how I think it basically works, based on reading bits and pieces on the web:
So – quick, but doesn’t (afaIk) take base-call qualities into account, can’t cope with more than 2 mismatches and can only deal with Illumina data (and Illumina length reads). Not sure how ELAND-extended differs. My guess is that the alignment to the first 32 bases is exactly the same, then they do some post-hoc aligning of the extra bits of the sequence?