You can then use
reformat.shfrom BBMap suite to filter your data where the merged read is exactly the same length as R1/R2 (I am assuming your reads are all identical length to begin with and have not been trimmed, e.g. you could set
n = length of R1/R2). That will filter out all reads where R1 and R2 perfectly match (your requirement).
If R1/R2 perfectly match but have a shorter insert than the length of sequencing, those reads would also be removed by filter above.