As someone also said, one of the biggest enemies of PCR is contamination, or perhaps the samples had conservation / handling problems which meant the RNA ended up degraded.
Given the (currently) very low numbers of these "apparently relapsed" cases versus the total recovered cases, I think it is most likely.
- False negatives, and so people were still actually sick
In addition, I wonder if those patients had distinct clinical features compared to others. In absence of more data (and these cases are very uncommon) it is hard to tell.
Probably a good idea would be to do a study with periodical follow up of a large cohort of discharged patients. This would probably give better indication.
Given the (currently) very low numbers of these "apparently relapsed" cases versus the total recovered cases, I think it is most likely.