Real-time PCR is quite good at detecting Ebola, but it's not 100%, and I think people are looking for 100% when it comes to Ebola.

If a person has Ebola, then PCR will most likely detect it, unless the viral load is too low to detect in the blood. Someone who is infected but not yet symptomatic could have too low a viral load to detect. Therefore, a single negative PCR results may be erroneous, which is why repeat tests after a couple days are recommended.

But, if a person who is infected has a viral load too low to detect with PCR, they are not going to be infectious anyway. It is best to run the test against people who recently came from an Ebola region or had contact with an Ebola patient and is starting to show symptoms. Actually, even after symptoms appear, the viral load can still be too low for the first 48 hours.

For reference, here's a flowchart for determining whether an Ebola case is "laboratory confirmed." (This is a European protocol but I believe the CDC uses something similar)