The options read2pos and readExtension3 are usefull for counting fixed fragments (Feature request for Rsubread::featureCounts: read length adjustment). Recently I considered one condition of counting reads based on their 5' most base, whose position was shifted. For example, the 5' most base is first shifted by a fixed number based on POS and CIGAR fields, then reads are summarized. It is useful if the middle point of fragments are considered in some IP experiments. So an added option shift and read2pos will be able to handle this situation well.
And, is it possible to add another option to count reads only in plus or minus strand of the genomes?
I also expect these options can be added in the linux version.
You can use read2pos=5 together with readExtension5 to shift the 5' most base of a read to any position you want. The read is first extended upstream by number of bases specified by readExtension5, and then it is reduced to its new 5' most base. Note that the 5' most base can be shifted downstream as well if you provide a negative value to readExtension5. Would this enable you to do what you want to do?
Regarding your request for counting positive-strand only or negative-strand only reads, I think you can simply subset the counts by performing filtering on the strandness?