Does anyone have any experience extracting unmapped reads from BAM file and saving them in a new BAM file using Rsamtools (or any other R/Bioconductor package) ?

I am analysing paired-end RNAseq data processed by Rsubread package.

If you could provide some generic code to address this problem it would be greatly appreciated.

Use filterBam() to create a new file from a subset of the original file, using the ScanBamParam() argument to select just unmapped reads. The help page ?filterBam includes this example

     param <- ScanBamParam(
        flag=scanBamFlag(isUnmappedQuery=FALSE),
        what="seq")
     dest <- filterBam(fl, tempfile(), param=param)
     countBam(dest)  ## 3271 records

(the flags can be complicated, so verify that this selects the reads you're interested in).