bustools workflow for snRNA-seq detects many more genes/UMIs than cellranger

[jrknoedler]

I'm a big fan of kallisto for bulk RNA-seq so decided to establish a kallisto | bustools workflow for my single nucleus RNA-seq data (admittedly partially because my university's cluster start throwing OOM-kill events when I was running cellranger, which had worked fine previously, for no obvious reason). I followed the tutorials for building a cDNA index by downloading the files directly from the Pachter lab's github repository and running the kallisto bus allignment, then bustools capture and count for the intron and cDNA alignments following the tutorial. So far, so good. However, there's an odd result - I'm getting something like 30% more unique genes/cell with Kallisto/bus than with cellranger-3.1 (something like 4,500 versus 3,000). Now, I don't want to complain, especially since this has already resulted in my being able to find markers that previously weren't showing up, but I also go by the credo that if something sounds too good to be true it usually is. Is this likely to be due to pseudoalignment to a cDNA/intron index performing better than cellranger? Has anyone seen anything like this before?