[marcosdumay]

It's not a problem with the theory, because the theory was not created to be a high-performance application.

It's also clearly not a problem with SQL since people very loudly and clearly prefer the version with nulls and operations that create nulls.

It is a problem with the idea that people should program in relational algebra. They shouldn't. It's the same kind of issue as functional languages that extend lambda calculus; logical languages with explicit evaluation strategies; type systems with undecidable situations; module systems that allows access to private features... No actual usable system is a perfect representation of the math behind it.