A method and apparatus are provided for identifying a material with a sample-specific reference spectral list or library. A sequential approach to SEM-EDS automated mineralogy classification is carried out by performing two or more material classification analyses. A pre-classification step restricts the processing of spectra deconvolution algorithms to a subset of spectra that pass a dominant mineral criteria resulting in a significantly reduced subset of reference spectra that occur within the measured sample in pure enough form at a given minimum quantity. The following complex classification stages involving deconvolution of multiple constituents within measured spectra is based on this sample relevant subset.