Single cell RNA-seq (scRNA-seq), which provides a snapshot of gene expression on a cell-by-cell level, is a method used to study heterogeneous tissues and discover new cell types, such as in analyzing the composition of cancerous tumors and classifying cell types in the brain. While it is known that genes can have alternative splicing isoforms that are important for different biological functions and contexts, the use of gene quantifications in computational analysis discards all information about isoforms. Utilizing pseudoalignment and transcript compatibility counts, a data structure that preserves isoform information present in scRNA-seq, we present methods for clustering and performing cell marker discovery that reveal previously unidentified cell types. In the 10x dataset of human peripheral blood mononuclear cells, we differentially clustered naïve and memory T cells, obtaining for these clusters markers CD45RA and CD45RO, respectively, which were previously thought to be undistinguishable using only 3’ end scRNA-seq. As shown by this analysis, by using transcript compatibility counts our methods do not require transcript quantification and are applicable even to non-uniform sequencing datasets, allowing the extraction of isoform-related information without full length sequencing.  As such, our methods are fast and scalable to large, sequencing-biased datasets with tens of thousands of cells.