2) I used Prodigal in the meta mode, because I am dealing with metatranscriptomes (bacterial community). I don't need to distinguish ncRNAs or 5' and 3' UTRs, I don't think. But, I thought I should define ORFs if I am interested in particular transcripts. Is this not so? Are you asking because there is a better ORF caller for my purposes, or because it is unecessary?
I based the decision to use Prodigal off of this Davids et al 2016 paper: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0146423
3) I should be more specific. I am ultimately interested in transcript counts of particular genes, because I want to see if there are genes or types of genes being expressed more in particular treatments compared to control treatments. So, I don't think what you described is the statistic I want.
Don't supply output names to samtools index command, only input name.