A version of the hg19 genome was selected as the target genome for the failed jobs. These failed jobs also made use of the same hg38-based reference annotation GTF (dataset 9) that was used for the successful jobs mapped against hg38. Try running this failed job against hg38 instead as it appears that is what you intended to do based on the other content of the same history.
If you really did want to map against hg19, then you will need to obtain a reference annotation dataset from UCSC that is based on hg19 and it that for the reference annotation ("known splice sites").
All inputs must be a match for tools to interpret the data correctly. Jobs with mismatched inputs will usually fail but some may produce data that is scientifically invalid. If you ever want to double check what settings were used for a job, the "double-circle" rerun button brings up the tool form view and the "i" info data details button brings up a summary report view.
Hope this helps! Jen, Galaxy team