1. You have two contigs with the same ID in your assembly (one longer and one shorter).
2. Your BioPerl index for retrieving the sequence is corrupt somehow. Requires delete of previous output directory before restarting.
3. You are using bad formatted GFF3 as input into maker, and it is somehow not failing right away.
The fact you get output means that it was able to translate the sequence into protein with CDS etc. So it was not too short for that. Look at the contig line in the GFF3 to get the length of the contig maker is seeing to compare to the feature positions given.