To my best knowledge, the noise reduction tools included in Photoshop apply the same filter (kind of low pass filter) in the entire image, regardless the local contrast. As a consequence, you cannot expect to go very high in the ratio noise reduced/details lost.
The dedicated noise-reduction software have totally different math inside. The key word is "anisotropic diffusion PDE" (PDE stands for Partial Differential Equation). In a word, it acts like a blur brush with the size and, moreover, the shape of the brush continuously adapted to the local gradients of the image.
PDE-based image processing is a very active field in Applied Math Research.
But dont expect too much from these magic (and comutationally intensive) math unless you end up with uggly plastic skins beyond salvation.
I dont use it any longer. Some grain in the image is very well accepted by most people when they know the lighting conditions are bad (in my case indoor sports at any event my children attend to). If the grain is so bad that people will not accept it, Its beyond salvation and Noise Ninja just get thing worst.