I found myself to force to use cellranger. Meanwhile it helps a lot to run from bcl files to single cell counts matrixes, I discovered that is quite difficult to control many options related to optimization.
I have to run more than 200 samples in a short time of period. In my current position at MIT, I joined the OpenMind cluster in the McGovern institute. I was pleased to find a very flexible cluster, but as any other cluster you need to respect other users.