I bet that the design is telltale of the computational power that was available for simulations at the time they were designed. They probably had to use clunky

I bet that the design is telltale of the computational power that was available for simulations at the time they were designed. They probably had to use clunky approximations for the B1, and the resulting design was clunky as a result. And better technology and knowledge of aerodynamics let them design models with smoother (read: more computationally intensive) contours later. Whether these smooth shapes were achievable in the 60s is a difficult question. They had parametric curves back then, but probably less idea how these contours influence aerodynamics and radar cross section, and even much less what the optimal shapes would be.