I have a feeling that the basic "need" for ORMs was originally driven purely by the unwieldiness of querying a MySQL database. When everything is in the application layer, you have to write abstractions on top of it to keep it comprehensible.
When you have triggers and views, on the other hand, you can just "ask the right questions" of the database, and get your information; the SQL becomes self-documenting, rather than a tangled mess.
Whatever their merits, ORM layers and the problems they try to address existed long before and quite independently of the particulars and limitations MySQL.
I've always felt that begs the question, by assuming that a mismatch is somehow fundamental, rather than an implementation artifact.
Just as real RDBMSes don't strictly adhere to the formal math relational model, object-oriented languages vary in their, for lack of a better word, orientation.
Looked at another way, what about a relational-object model instead? Making a useful one for MySQL doesn't seem as plausible, given its chronic feature absence.
When you have triggers and views, on the other hand, you can just "ask the right questions" of the database, and get your information; the SQL becomes self-documenting, rather than a tangled mess.