In 2006 in Critical Care Medicine it was shown to likely be superior to WBC and CRP for predicting mortality, however, when studies again in 2010 by Critical Care Medicine in a meta-analysis of 7 randomized clinical control trials, it did not demonstrate any difference in mortality, however, did results in overall 4 days less antibiotic use on average. The journal Shock in 2008 published a small study in a non - U.S. site which suggested that procalcitonin was superior to CRP for predicting mortality. The FDA approved its use based on this when used in conjunction with other indicators at a cutoff of < 0.5 being low risk, and > 2 being high risk.
The Cochrane collaboration in 2012 showed that there was again no different when procalcitonin was used and the confidence intervals demonstrated to statistical significance.
Hopefully we will have a magic test in the future that will clearly distinguish between bacterial and viral sources of infection, however, in the meantime... keep using your clinical gestalt.