Is it normal to get different counts for the same gene when transcript version numbers are removed vs when they are left in the tx2gene matrix? How does tximport handle these two different situations?
You will have to provide more information than that if you want help. Although if I assume you are using Ensembl identifiers, it can be tricky to find the right EnsDb to do the mapping if you are working on versioned transcripts. Unless all the versioned transcript IDs are in your EnsDb you won't get the same mappings with and without the versions.
Reminder that tximeta exists -- if you are working with mouse or human data and using Ensembl or GENCODE (or RefSeq, but not "latest" which isn't anyway a stable version). It will take care of all of this on behalf of the user.
You will have to provide more information than that if you want help. Although if I assume you are using Ensembl identifiers, it can be tricky to find the right
EnsDbto do the mapping if you are working on versioned transcripts. Unless all the versioned transcript IDs are in yourEnsDbyou won't get the same mappings with and without the versions.