GTF2/GFF3 "feature" types and expression analysis

Hi, I aligned a few samples using STAR to the genome provided in the Illumina iGenomes UCSC hg19 bundle (here) -- I used the provided gene feature (gtf2) file as is.  Now, my motive is to calculate the gene and isoform expression levels using bedtools multicov (at the same time). Use of the gtf2 file produces a file containing read counts per exon.  I wish to compute gene and isoform read counts too, so I converted the gtf2 file to a gff3 file using using gtf2gff3 script from SO/GAL (here).  My first question is: Is it OK if the alignment is performed with gtf2 file but counted for reads using the gff3 file, keeping in mind that the gff3 file was converted from the gtf2 file? My second question follows I have read both these resources (here and here) but do not understand the differences between:

• exon vs CDS
• transcript vs mRNA

I know that with the process I described, it is possible to retrieve gene read count by selecting only the lines where feature=gene from the bedtools multicov output.  What must I do for isoforms?  I am confused by the semantics. Thanks ahead of time and let me know if my post was not clear enough. ...