It sounds like you are assembling mRNA-seq reads into transcripts and trying to align these to produce gene models?

If so, I typically would run exonerate like: "exonerate -m est2genome --revcomp --bestn 1 --showcigar --showtargetgff -t chloroplast.fa -q scaffolds.fa > scaffolds.out 2> scaffolds.err", then extract the GFF lines from scaffolds.out. However, this will not generate CDS features in the GFF.

If CDS annotations are important, then first run ORF prediction on your sequences and produce a 4-column, space-delim file ("CDS.txt"), one row per scaffold, containing these 4 values: scaffold ID, strand of ORF (+ or -), scaffold ORF start (1-based), scaffold ORF end. With this file, you can run exonerate like this: "exonerate -m cdna2genome --annotation CDS.txt --revcomp --bestn 1 --showcigar --showtargetgff -t chloroplast.fa -q scaffolds.fa > scaffolds.out 2> scaffolds.err". The GFF will now include CDS lines.

Depending on the results, you may want to change default values for --refine, --minintron, --maxintron. Exonerate is a parameter jungle so you may find other useful ones, but these are what I typically use.