RGI bwt: paired end VS single end. Which one is better?

[valei]

Hi, I ran RGI bwt in a metagenome sample in two different ways: 1) using as input R1 and R2 paired-end files; 2) using as input a concatenated file R1+R2 and so considering the sample as single-end. The results at both allele and gene level are fully different for a large number of genes/alleles found (about 30/40%). I wonder, why is there such a big difference? What is the most reilable way to run the tool?