manta_run_workflow fails with Unexpected character in reference sequence.

[Parsoa]

I get the following error during the manta_run_workflow stage:

[2020-05-17T19:56:16.703129Z] [23217_1] [WorkflowRunner] [ERROR] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000] Error Message:
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000] Last 6 stderr lines from task (of 6 total lines):
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [2020-05-17T19:53:07.332513Z] [23217_1] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000] ERROR::
Unexpected character in reference sequence.
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [2020-05-17T19:53:07.334080Z] [23217_1] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000]   refer$
nce_sequence_file: '/share/data/BayesTyper/bayestyper_GRCh38_bundle_v1.3/GRCh38.fa'
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [2020-05-17T19:53:07.346773Z] [23217_1] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000]   chrom$
some: 'chr19_KI270933v1_alt'
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [2020-05-17T19:53:07.347916Z] [23217_1] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000]   chara$
ter: '>'
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [2020-05-17T19:53:07.348516Z] [23217_1] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000]   chara$
ter_decimal_index: 62
[2020-05-17T19:56:16.705237Z] [23217_1] [WorkflowRunner] [ERROR] [2020-05-17T19:53:07.349091Z] [23217_1] [makeLocusGraph_chromId_438_chr19_KI270933v1_alt_0000]   chara$
ter_position_in_chromosome: 118438
gzip: manta/NA19240/results/variants

I get a similar error for all makeLocusGraph workflows launched for alt contigs. The stage eventually fails with the following error:

gzip: manta/NA19240/results/variants/candidateSV.vcf.gz: No such file or directory
gzip: manta/NA19240/results/variants/candidateSV.vcf: No such file or directory

As for this particular case, I checked the sequence for chr19_KI270933v1_alt in the reference file provided by the BayesTyper bundle just in case and it doesn't include any > characters in the reported position. I'm assuming this stage does not look at the BAM file yet, so it shouldn't be a problem with the mapping. Any insights on this? Thanks.