That has a unique solution, so it is infinitely better than the linear error, but is not a nice differentiable signal suitable for controls, I'd bet. But that wasn't in the problem statement.