tophat v2.0.7 segment-based junction search error -11

[mmuratet]

Greetings
I am using a set of scripts to submit and run tophat jobs that has worked well in the past. Lately I have seen runs crash with errors that I am having a hard time connecting to data or parameters, e.g.,
[2013-06-24 11:48:49] Beginning TopHat run (v2.0.7)
-----------------------------------------------
[2013-06-24 11:48:50] Checking for Bowtie
Bowtie version: 2.0.6.0
[2013-06-24 11:48:51] Checking for Samtools
Samtools version: 0.1.14.0
[2013-06-24 11:48:51] Checking for Bowtie index files
[2013-06-24 11:48:51] Checking for reference FASTA file
[2013-06-24 11:48:51] Generating SAM header for /gpfs/gpfs1/reference/genomes/bowtie2/mm9
format: fastq
quality scale: phred33 (default)
[2013-06-24 11:50:35] Reading known junctions from GTF file
[2013-06-24 11:50:54] Preparing reads
left reads: min. length=100, max. length=100, 35974176 kept reads (14522 discarded)
right reads: min. length=100, max. length=100, 35481156 kept reads (507542 discarded)
[2013-06-24 12:42:03] Creating transcriptome data files..
[2013-06-24 12:43:41] Building Bowtie index from Mus_musculus.NCBIM37.67.fa
[2013-06-24 13:01:37] Mapping left_kept_reads to transcriptome Mus_musculus.NCBIM37.67 with Bowtie2
[2013-06-24 13:53:29] Mapping right_kept_reads to transcriptome Mus_musculus.NCBIM37.67 with Bowtie2
[2013-06-24 14:45:06] Resuming TopHat pipeline with unmapped reads
[2013-06-24 14:45:07] Mapping left_kept_reads.m2g_um to genome mm9 with Bowtie2
[2013-06-24 15:39:31] Mapping left_kept_reads.m2g_um_seg1 to genome mm9 with Bowtie2 (1/4)
[2013-06-24 17:04:06] Mapping left_kept_reads.m2g_um_seg2 to genome mm9 with Bowtie2 (2/4)
[2013-06-24 18:27:30] Mapping left_kept_reads.m2g_um_seg3 to genome mm9 with Bowtie2 (3/4)
[2013-06-24 20:07:54] Mapping left_kept_reads.m2g_um_seg4 to genome mm9 with Bowtie2 (4/4)
[2013-06-24 21:43:09] Mapping right_kept_reads.m2g_um to genome mm9 with Bowtie2
[2013-06-24 22:38:26] Mapping right_kept_reads.m2g_um_seg1 to genome mm9 with Bowtie2 (1/4)
[2013-06-25 00:00:36] Mapping right_kept_reads.m2g_um_seg2 to genome mm9 with Bowtie2 (2/4)
[2013-06-25 01:27:19] Mapping right_kept_reads.m2g_um_seg3 to genome mm9 with Bowtie2 (3/4)
[2013-06-25 02:53:27] Mapping right_kept_reads.m2g_um_seg4 to genome mm9 with Bowtie2 (4/4)
[2013-06-25 04:19:13] Searching for junctions via segment mapping
[FAILED]
Error: segment-based junction search failed with err =-11
BAM read: invalid CIGAR operation

Here are the parameters I used,
params="-p 8 --no-coverage-search"

I checked all the bam files in /tmp directory with the picard ValidateSamFile tool. I see errors about "out of coordinate order" in files tophat wrote. I checked the input fastq files and they appear to be in the same order and properly formed for the first 10 records at least. I see nothing abouT CIGAR in the ValidateSamFile output

We recently changed schedulers from moab to lsf. I checked to be sure that the submission command restricted the job to a single node. It does.

I'm at a loss for what to check next. Anybody have any suggestions?
Thanks
Mike