We have not had a lot of luck with the reverse reads using the Illumina ITS1 primers from Dylan and Kabir (Smith and Peay 2014, PLoS ONE). As such, we have just been using the forward reads (R1) in the NGS analyses we have been doing. Pairing seems like it could be great, but see Nguyen et al. 2015 for empirical evidence that pairing did not help improve recovery of the mock community. I think that others found similar patterns, but unfortunately I can’t recall the specific paper where I saw that reverse reads faired more poorly than forward reads.