In constructing personas, we had to be cognizant of inadvertently creating stereotypes as humans naturally stereotype as a way of categorizing conceptions of others
In addition to this inadvertent tendency to create stereotypes, I think that in only making a couple of personas to represent the learning audience you may fall into stereotyping just by lack of a sufficient sampling. How do you determine how many personas would be a representative enough sample? If you are looking at a diverse group of learners, you need more personas, and you need to have instructional materials that cover diverse needs. In larger groups would you break the group into sections to better address individual needs? Or have additional instructors?