This is a short note, collecting thoughts on checking for consistency.

We have observed that using ELAND for alignment, we have an offset
error for reads on the reverse strand. For a yeast example, ELAND was
run using only 25 bases (this is choosen at compile time) and the
positions reported for hits on the reverse strand have to shifted
(subtracted) 11 in order to be "correct".

Here correct means that two reads at the same position and chromosome,
but with different strand, lies opposite each other. This is easiest,
but may not be the best representation since the base quality
decreases.

I checked it by taken a U0 hits on the reverse strand and ran BLAT
from the UCSC genome browser. It can be a bit hard to figure out which
strand you hit, but click "details" (and not "browser") and you get
the information (ie. strand and start). The UCSC genome browser report
position relative to the 5' end of the forward strand.

This might be automated if there is a "quick" way to see if we have
two reads with the same position and chr, but different strand.

A BLAT example: I get 
  chr6:57880-57915, reverse strand
from BLAT, whereas I have
  chr06.fsa 57891 R
from ELAND. The conclusion is that I need to subtract 11 from the
  ELAND reported positions.
