I am one of the featureCounts authors.
Do you want do conduct analyses for genes, or do you need individual results for all the isoform variations of each gene? The "isoforms" tend to be called "transcripts" in the RNA-seq literature. The featureCounts help page you link to is designed to generate gene-level counts. We have never recommended the use of featureCounts to produce transcript-level counts.
A gene level analysis is easiest to interpret and is appropriate for most analyses where you want to interpret the results in terms of pathways, GO terms, KEGG and so on. featureCounts will give you gene level counts very quickly and easily. Alternatively, featureCounts can give you counts for each exon in each gene.
If you do really want to quantify all the different transcripts (isoforms) for each gene, then this cannot be done by featureCounts or by any other read counting software. While featureCounts can be run with transcript annotation, the results won't be particularly useful. Since the transcripts are heavily overlapping, with many transcripts sharing the same exons, featureCounts cannot tell which transcript to assign each read to. It therefore has to either discard all the multi-assigned reads or alternatively assign them multiple times to all the overlapping transcripts -- and either choice is bad.
If you really do want the (considerable) extra complexity of transcript-level analysis, then I would run Salmon and read the results into edgeR using
catchSalmon. Salmon can assign reads to transcripts on a probabilistic basis. Or alternatively, do an exon-level analysis. That allows you see what is going on, although the results are not assembled into transcripts. Exon-level may be a safer approach if your data has expressed transcripts that have not been previously annotated.