bypassing: "MESSAGE: Query asks for data past end of contig" error?

Hi,

I'm running the following command as part of the rna-seq based variant detection pipeline:

java -jar GenomeAnalysisTK.jar -T SplitNCigarReads -R target.fasta -I dedupped.bam -o splitNCigar.bam -rf ReassignOneMappingQuality -RMQF 255 -RMQT 60 -U ALLOW_N_CIGAR_READS --validation_strictness LENIENT

and it throws a fatal error:

  ERROR AN INPUT FILE ERROR has occurred (version 3.8-0-ge9d806836):
  ERROR MESSAGE: Query asks for data past end of contig

Is there a way to have it bypass alignment records that are malformed in some way?