See attached article and references cited therein explaining why even the PMF and UNMIX solutions to the CMB equations require comparison of the source factors with real-world source profiles.
ARAI (2009). Source profiling for vehicular emissions. Report Number ARAI/VSP-III/SP/RD/08-09/60 ; prepared by Autmotive Research Association of India, Pune, India, http://www.cpcb.nic.in/Source_Profile_Vehicles.pdf.
Sethi, V.; Patil, R.S. (2008). Development of Air Pollution Source Profiles - Stationary Sources Volume 1. prepared by Indian Institute of Technology, Mumbai, India, http://www.cpcb.nic.in/Source_Emission_%20Profiles_NVS_Volume%20One.pdf.
Sethi, V.; Patil, R.S. (2008). Development of Air Pollution Source Profiles - Stationary Sources Volume 2. prepared by Indian Institute of Technology, Mumbai, India, http://www.cpcb.nic.in/Source_Emission_Profiles_NVS_Volume%20Two.pdf.
It means that its always better to built comprehensive emission inventories for individual cities with spatially and temporally resolved real samples to get authentic source-apportionment, am I right?