You need to create a custom reference GTF where each of your exon is a gene. So a uniq gene_id and gene_name for each of your exons and then use that GTF in cell-ranger pipeline from the beginning.

In GTF file, the gene_id and gene_name attributes are used to get gene level counts, which basically sums up exon level counts that belong to a same gene. if you want exon level counts, each exon should have its own unique gene_id gene_name attribute.

For example, lets say you have a gene definition ( dummy example):

chr1 dummy  exon          380   401   .   +   0  gene_id "001"; gene_name "aa";
chr1 dummy  exon          501   650   .   +   2  gene_id "001"; gene_name "aa";
chr1 dummy  exon          700   707   .   +   2  gene_id "001"; gene_name "aa";
chr1 dummy  exon  380   382   .   +   0  gene_id "001"; gene_name "aa";
chr1 dummy  exon   708   710   .   +   0  gene_id "001"; gene_name "aa";

That should be changed to:

chr1 dummy  exon          380   401   .   +   0  gene_id "001_ex1"; gene_name "aa_ex1";
chr1 dummy  exon          501   650   .   +   2  gene_id "001_ex2"; gene_name "aa_ex2";
chr1 dummy  exon          700   707   .   +   2  gene_id "001_ex3"; gene_name "aa_ex3";
chr1 dummy  exon  380   382   .   +   0  gene_id "001_ex4"; gene_name "aa_ex4";
chr1 dummy  exon   708   710   .   +   0  gene_id "001_ex5"; gene_name "aa_ex5";

Then instead of getting counts for gene 001 , now you will get counts for each individual exons in your output.

Note: The exons need not to be renamed to 1,2,3,4, etc. They can be given any unique value.