The test for something like testosterone will be very straightforward, so by splitting the sample in two you’re essentially eliminating false positives for practical purposes.
With these natural substances that are present endogenously, there’s still always going to be some amount of wriggle room for the athlete to try and bullshit their way out of it. You’re testing for an unusual ratio of compounds that are present naturally.
For a synthetic steroid, say stanozolol (winstrol), no amount of bullshitting can save the athlete. (Interesting conspiracy theory - Ben Johnson got done for Winstrol and claimed his sample was spiked by another athlete. He freely admits to being on the gear, as was every other sprinter in the Seoul 100m final, but said he never took Winstrol as it made his joints feel tight).
A false positive will be more likely for a more complicated test - say when the first EPO tests were being developed, you’re trying to distinguish subtle differences between the natural protein and the recombinant one, which I’d expect to be more challenging and more subject to errors. The UCI just placed a limit on red blood cell count in the interim, (actually still in place I believe), whilst the EPO test was developed to a reliable standard.