The point is not that it's a definitive test, the point that it's a reasonably accurate
blood and urine test. As in, after discussing recent problems with your doctor, your doctor may then conclude that this would be a good time to stick you with a biopsy needle and test for pancreatic cancer.
But wait, this is invasive and potentially harmful, is there some way we can be a bit more sure about things before we confirm?
Why yes! This kid developed a blood and urine test which is 90% accurate!
The point is to potentially reduce the number of large, expensive needles stuck into someone's pancreas, not to serve as a standalone test.
It also matters WHY the test is inaccurate. If it's consistent with each individual "if I get a false positive, it will ALWAYS be a false positive" because of a lack of a certain protein or whatever, then it's less useful (unless you determine the conditions that make it work). If it's actually just a random 10% due to lack of precision for a particular measurement, then it can be refined, OR you could just run it five times and do some math to get a result with >90% accuracy.