Well letters don't really have a single canonical shape. There are many acceptable ways of rendering each. While two letters might usually look the same, it is very possible that some shape could be acceptable for one but not the other. So, it makes sense to distinguish between them in binary representation. That allows the interpreting software to determine if it cares about the difference or not.

Also, the Unicode code tables do mention which characters look (nearly) identical, so it's definitely possible to make a program interpret something like a Greek question mark the same as a semicolon. I guess it's just that no one has bothered, since it's such a rare edge case.