Problem with BWA mapping second read

[pbluescript]

I am mapping some 2X100PE RNA-Seq data and BWA is mapping the second read very oddly for some of my samples.
I am testing out a mix of BWA and Tophat for mapping the reads and sometimes BWA will say two reads in a pair map to the same location even though the second read is filled with mismatches. I have attached an example IGV screenshot showing the BWA+Tophat mapped reads on the top and the Tophat only mapped reads on the bottom.
This happens for some of the samples, but not for others, even though the libraries were all prepared the same way. The coverage graphs on the top are identical, and since it is FPKM, there isn't likely to be much effect when it comes to differential expression analysis. Indeed, when comparing the sample mentioned above mapped with either BWA+Tophat or just Tophat, the R-squared is 0.9705.
I was planning on playing around with the BWA settings a bit, but I thought I'd pose the problem here to see if anyone had some suggestions.

[prm36]

I find that by running bwa sampe with the -s option, the second read of the pair does not get mapped, and thus resolves this issue.

[prm36]

I just checked my lab book for my notes for when I found myself having the same problem.

"""
Reads with multiple mismatches are mapping.
If one read of a pair maps, SW algorithm in bwa will map pair regardless whether paired read passes mismatch filter or not.
Workaround
Run bwa aln with -n 1 (allow 1 mismatch per read)
Run bwa sampe with -s (disable SW algorithm)
"""

[pbluescript]

Thanks for the advice prm36. Unfortunately, it didn't seem to help. I get identical numbers of reads mapping and the same problem when I use the -s option with bwa.

Any other suggestions?