The decision of whether to include multi-mapping reads or not depends on the specific research question and the downstream analysis. Multi-mapping reads are reads that can align to multiple locations in the genome, while unique reads can only be aligned to a single location. In general, including multi-mapping reads can increase the number of reads that are counted, but it can also increase the likelihood of counting reads from paralogs or repetitive regions.

If your research question is focused on detecting alternative splicing or identifying genes that are expressed at low levels, then including multi-mapping reads may be important to capture all possible expression events. However, if your goal is to identify differentially expressed genes between two groups, then it is generally recommended to exclude multi-mapping reads to reduce the potential for false positives.

It is also important to note that different aligners and quantification tools may handle multi-mapping reads differently, so it is important to carefully consider the options available in your specific analysis pipeline. Ultimately, it is recommended to perform sensitivity analyses with both options (including and excluding multi-mapping reads) to determine which approach gives the most biologically meaningful results for your specific research question.