The > and < are reference skip symbols and do not (directly) have any particular exon/intron interpretation. They are described in the samtools manual in the paragraph starting "In the pileup format...". The quality score encoding is described there too. The question titled Some help understanding with mpileup output also discusses the mpileup format.

Thanks for pointing to the "reference skip" definition. I still have fully understood the "quality score" aspect of some reference skips. Here is my question. I am looking at mpileup output from RNA-Seq data from one sample. And the pileup output is something like

chr1 3203517 T 30 <<<<<<<<>>>>>><><>>>>>><><><<< IIIIIIHIE@HFGHIFIIIGHHIIDIHHGD

I also looked at the location in IGV and found that the location is intronic. All the reads that map at the location covers the two exons adjacent. Here is a toy example of the scenario, showing three reads that spans two exons.

                       |<- location of interest
               EXON1-------EXON2
  R1              AT-------TAG
  R2            ATAT-------TA
  R3              AT-------TAGA

Basically, no real bases are at the location, but mpileup gives quality scores for the "bases".

Does mpileup come up with random quality scores just to keep the format of mpileup intact?