cellranger count needs files in a special format. There are three separate files (if you have > 1 sample). R1 file contains the cell barcodes and UMI, I1 file contains Illumina index for the sample, R2 file contains actual reads. It appears that these investigators may have submitted these fastq files in a concatenated format. Try to see if you can separate the reads using --split-files option for fastq-dump.

I have the same issue, did the split-files solved your problem ?

Just my two cents here. In my case cellranger count was failing because R1 was in the wrong format. I also had single-end reads, where R1 was supposed to be the cell barcode and UMI, and R2 should be the single-end reads. In my case I had to rerun cellranger mkfastq from the raw BCLs. Note that the I1 file mentioned is optional for cellranger count so don't worry if you don't have it.

In my experience the --split-files option for fastq-dump is only relevant if you're downloading data from SRA, not if you have your raw data in-house.