Contamination between samples run on the same lane

[MLog]

Hi colleagues,
I found a strange effect when sequencing several samples on Miseq. The reads from each sample contain about ~0.1% of reads from other samples run on the same flowcell. This does not seem to be caused by improper recognition of barcodes - we did demultiplexing without allowing mismatches, and the indices are very different one from another (e.g. TAGCTT and AGTTCC).
And this is certainly not contamination during library preparation - because I had one of the samples sequenced twice, with different "neighbors" and the contaminating reads were different.
Have you ever seen something like that? What can be done to minimize this?

[GenoMax]

See if your observation is the same as recently mentioned here: http://seqanswers.com/forums/showthread.php?t=38697

[MLog]

No, this is not a contamination between runs, if you mean this.

[GenoMax]

Check the posts by vl80 and JackieBadger in the thread I linked above. They are about cross contamination between samples in the same run. That is what you are referring to right?

[mcnelson.phd]

Yeah, it looks like you're seeing what I've been calling index misassignment. It apparently stems from image alignment and signal bleed issues during index sequencing that causes a cluster to be assigned the wrong index sequence.

Unfortunately, nothing can really be done about it, although dual indexing will cut down on the misassignment issue.