Transcripts not found when building database for custom genome

[mrbertp]

Describe the bug
I am trying to build a snpEff database for my custom genome (S. japonicus, fission yeast, assembly SJ5). I have all genome.fa, cds.fa and protein.fa files ready and checked that their ID match the transcript ID, and also added the genome to the snpEff.config file. When I build the database (java command below) a lot of warnings about transcripts not being found are displayed (see below). Also, the sanity checks for CDS and protein sequences fail with >35-37% of error.

I tried to reproduce the example of database building for the human genome (https://pcingola.github.io/SnpEff/snpeff/build_db/#example-building-the-human-genome-database) and obtained a similar result (trancripts not found).

To Reproduce

1. SnpEff version: 5.3a (build 2025-09-02 10:24)
2. Genome version: SJ5
3. SnpEff full command line: sudo java -jar snpEff.jar build -gtf22 -v SJ5 2>&1 | tee SJ5.build
4. Output / Error message:
   SJ5.build.txt

Expected behavior
I should be able to build a database with <3% of errors.

Data
Annotation files (cds.fa, genes.gtf, protein.fa):
annotation.zip
Genome:
SJ5.zip

Additional context

[mrbertp]

I do not know if this has to do with the problem, but if you notice, within each WARNING_TRANSCRIPT_NOT_FOUND block in the SJ5.build file, the coordinates for the transcript created:

I AGAT EXON 21840 23456

are 1 position ahead of the coordinates of the pre-existent feature read:

I AGAT exon 21841 23457

Full warning block:

WARNING_TRANSCRIPT_NOT_FOUND: Cannot find transcript 'SJAG_01096.1'. Created transcript 'SJAG_01096.1' and gene 'SJAG_01096' for I      AGAT    EXON    21840       23456   +
        gene_id : SJAG_01096
        id : agat-exon-2075
        parent : SJAG_01096.1
        transcript_id : SJAG_01096.1
. File '/home/bert/snpEff/./data/SJ5/genes.gtf' line 2  'I      AGAT    exon    21841   23457   .       +       .       gene_id "SJAG_01096"; transcript_id "SJAG_01096.1"; ID "agat-exon-2075"; Parent "SJAG_01096.1";'

[mrbertp]

If I use a GTF3 file as genes.gtf and I do not specify GTF flavour in the java command:

sudo java -jar snpEff.jar build -v SJ5 2>&1 | tee SJ5_gtf3.build

the warnings about transcripts not being found disappear.

The problem is that I still have 35-37% errors when it does the sanity check with CDS.fa and protein.fa.