Read on Twitter
PCR doesn't make RNA, it makes DNA (it even says so in the figure!). PCR isn't making "copies of a specific DNA/RNA sample", but of a sequence region (if present) in a sample. "Covid" isn't a virus.
Here's a warts-n-all example of the sort of "drift" you can get. The 2 sigmoidal (S-shaped) curves are a POS control & a clinical positive for HCoV-NL63 (this is data from 2010). The rest are negative samples-but there's some drift above the threshold as the test goes on.
This is a 4-target test and they can be tough to design & have work sometimes (the components can interact in unexpected ways). This "multiplex" test was abandoned but the assays still worked okay when running as 4 individual tests. You can just raise the threshold and "fix" it!
This RT-rPCR was run at out to 55 cycles. only the positive control (human parainfluenza 4 RNA) came up. Negative controls (water instead of clinical sample extracts) and clinical sample nucleic acid extracts all negative, despite long run
This was a rhinovirus RT-rPCR (real-time RT-PCR). Nicely shows how variable the results can be. These were from ill people, and while it cannot tell us that they were ill due to these viruses or that there is viable virus there-it does tell us they had been infected. Also,
it shows that most samples tested negative (including 2 negative controls) despite the cycles. The "late & low" pink curve is no as "pretty" as we'd like to see so it and the pain-in-the-arse L&L purple sample would need to be looked at further.
For these we could re-extract form eth original sample & repeat test (in case something went wrong with extraction), or just repeat the original extract (in case something went wrong with loading) or use a 2nd different test (different genetic region of virus)
Hopefully, you see that no, lab tests aren't always going to give clear-cut answers (biology is full of variation), but expertise-experience in using them & knowing the tests & thus what to expect-make the call simpler for the expert than for the non-expert.
Also, tests differ. And different labs-especially in some places-do their own thing. In Aus, we have some over-arching bodies that expect our tests to meet stringent criteria that include a lot of evaluation work before a test is "fit" for use in testing human samples
