NGS result that can't be verified with RT-PCR

[dlhmmll]

I received 5 plant leaf samples for NGS. Two of the 5 were determined to be positive for virus via RT-PCR. However when I performed NGS, all 5 samples were determined to be positive. In the 3 samples that were supposed to be negative, I acquired complete genomes when assembled using CLC Genomics Workbench.

I can't seem to find any source of contamination. The 3 supposed negative samples are of a different strain than the other two. However, it is a strain that should still be picked up by my RT-PCR primers.

I hadn't used these indices in a while as I know that I have had problems with false positives in the past when using the same index sequences in consecutive runs. This is why I always rotate my index sequences for each NGS run. The last several times these index sequences were used, they did not detect that target. So, I'm thinking that it's not that type of contamination.

I've ruled out that it's pipette contamination because I'm using the same pipettes on library prep as I do with RT-PCR.

The run was a near perfect 1452Kmm-2. So, overclustering leading to strange base calls can also be ruled out.

Does anyone have any ideas what could be going on here?