The characterization of respirable dust on the basis of constituent fractions and particle sizes is increasingly of concern for evaluating exposure hazards. For high-resolution particle analysis, scanning electron microscopy with energy dispersive X-ray (SEM-EDX) can be an effective tool. However, it requires particles to be deposited on a smooth, uniform substrate such as a polycarbonate (PC) filter for optimal results. While direct sampling onto PC is possible, this is not the standard approach in many situations. For example, in coal mines, respirable dust samples have typically been collected onto polyvinyl chloride (PVC) filters because they are intended for gravimetric and/or infrared spectroscopy analysis. Such fibrous substrates are not ideal for SEM-EDX (or other microscopy), but an effective method to recover and redeposit the dust particles could render such samples suitable for the additional analysis. Here, we present a simple method and compare SEM-EDX results for paired samples analyzed directly on PC and following recovery from PVC and redeposition on PC. Both laboratory-generated dust samples (n = 10 pairs) and field samples of respirable coal mine dust (n = 44 pairs) are included in this study. Although some changes in particle size distributions were observed between samples analyzed directly and those that were recovered and redeposited prior to analysis, the results indicate the dust recovery method generally yields a representative sample in terms of mineral constituents. That said, results also highlighted the effects of high particle loading density on individual particle analysis. Considering all sample pairs, those with similar loading density between the directly analyzed and recovered sample tended to exhibit similar mineralogy distributions. This was generally the case for the lab-generated sample pairs, and the Freeman-Halton exact test of independence indicated that the samples in just three (of 10) pairs were in disagreement in terms of their mineralogy distributions. On the other hand, for the field samples, the directly analyzed sample often had higher loading density than the recovered sample; and the Freeman-Halton test showed that 25 (of 44) pairs were in disagreement. However, the effect of possible particle agglomeration on the directly analyzed samples cannot be ruled out—and exploration of this factor was beyond the scope of the current study.