Where are the UTRs in Augustus?

[cascoamarillo]

OK, long story short; after training and prediction with WebAugustus, with "User set UTR prediction: true", I don't see them (UTRs) clearly in the gtf - gff files. Instead, you see this output (gff file example of one gene):

# start gene g1
Contig1	AUGUSTUS	gene	1	870	0.81	-	.	g1
Contig1	AUGUSTUS	transcript	1	870	0.81	-	.	g1.t1
Contig1	AUGUSTUS	tts	1	1	.	-	.	transcript_id "g1.t1"; gene_id "g1";
Contig1	AUGUSTUS	exon	1	870	.	-	.	transcript_id "g1.t1"; gene_id "g1";
Contig1	AUGUSTUS	stop_codon	432	434	.	-	0	transcript_id "g1.t1"; gene_id "g1";
Contig1	AUGUSTUS	single	432	854	0.95	-	0	transcript_id "g1.t1"; gene_id "g1";
Contig1	AUGUSTUS	CDS	432	854	0.95	-	0	transcript_id "g1.t1"; gene_id "g1";
Contig1	AUGUSTUS	start_codon	852	854	.	-	0	transcript_id "g1.t1"; gene_id "g1";
Contig1	AUGUSTUS	tss	870	870	.	-	.	transcript_id "g1.t1"; gene_id "g1";
# coding sequence = []
# protein sequence = []
# Evidence for and against this transcript:
# % of transcript supported by hints (any source): 33.3
# CDS exons: 0/1
# CDS introns: 0/0
# 5'UTR exons and introns: 0/1
# 3'UTR exons and introns: 1/1
#      E:   1 
# hint groups fully obeyed: 1
#      E:   1 (transcript49047)
# incompatible hint groups: 0
# end gene g1

So in this example, the gene "should" have an 3' UTR annotation, but there is no UTR on the 3rd column in the gene features.

Even more, if you go back at the output training files, and look for these cases (genes showing "# 5' - 3'UTR exons and introns: 1/1" in the prediction output), you see them with the UTR feature annotated (3'-UTR and/or 5'-UTR) in training/hints_utr_pred. Although same genomic coordinates, they have different gene number.

Since the forum in WebAugustus is for internal use only

http://bioinf.uni-greifswald.de/bioinf/forum

I am asking here.

So are those UTR predictions not robust enough? Or there is another extra step necessary to include them in the final annotation?
I'd appreciate any input on this issue. Thanks.