Just as an aside: "accuracy" is a pointless measure if you don't also know the ratio of true cases to false cases. The metrics you really want is the True/False Positive rates - without that you can't gauge the quality of a test like this.
The data I got is that among the people who really got the virus, positive rate of the test is only 30% - 50%. So false negative rate is as high as 50-70%.
The news report (in Chinese):