Why normalize coupling matrix to FBW?

Hello, I am going through "Microstrip Filters for RF Microwave Applications 2nd Edition" by Hong, and am not entirely satisfied by the explanation of coupled resonator circuits (chapter 7). When analyzing these circuits, they quickly dive into forming an impedance matrix, which makes sense. But the next step is to normalize the nxn impedance matrix thusly:

Where Zbar is the normalized matrix, from which relevant S parameters are eventually extracted. This is the first step towards deriving the nxn coupling matrix (and later the n+2xn+2 matrix).

I understand why it makes sense to normalize by removing wL, but why normalize by the fractional bandwidth (FBW)? If anyone can explain or has another source to direct me to which could provide better understanding, it would be much appreciated. Currently the only other good source I have found is Cameron"s "Microwave filters for communication systems : fundamentals, design, and applications".