Very low number of variants called by elprep in an heterozygous genome

[chklopp]

I'm trying to use elprep to call variants in a genome having one SNP every 50bp.
It runs but the g.vcf files have only hundred times less position where $5 != "<NON_REF>" as expected.
I use standard parameters in my command line without a reference vcf file because at this stage I would like to define the reference variation set by combining freebayes, bedtools and elprep results. freebayes and bedtools give high numbers of variants with most are concordant.
What should I modify in the standard elprep command line to increase the number of detected variants?