I don't think you can get any consensus on compensation design and crossover frequency. If you put 10 engineers in a room you will get 11 different answers. While both of the sources you cite are valid up to a point, I do not use either of them when I design compensation networks.
For example 1, the intent is to put the crossover frequency at the lower of the geometric mean of Fpmod and Fsw/2 or the geometric mean of Fpmod and the output cap ESR zero. That is all good but it assumes ideal current mode control with no slope compensation. When you add slope compensation, the equations become very complicated. But for this simple approach, the chosen crossover frequency is relatively low, so it results in a conservative design approach that can work for most circuits.
Example 2 only concerns with keeping the crossover well away from Fsw/2. Crossing over at or above Fsw/2 will violate the Nyquist criteria and is strictly prohibited. Choosing Fco between Fsw/4 and Fsw/10 is fine in theory, but the assumption here is an ideal error amplifier with infinite (or very large) GBW. For the TPS57112q1, I think you will find that there is not enough gain in the error amplifier to realize a 300 kHz crossover.
My preferred approach is outlined in the TPS54478 datasheet. I model the power stage gain and phase characteristics. I look for the frequency where the power stage gain falls to -120 degrees. That is the maximum Fco where 60 degrees of PM is possible with type 2 compensation. Depending on the slope of the phase curve, I may go a little lower or higher (higher if I have the option to add Cff to boost phase). Once I select the frequency I look at the gain of the power stage at that frequency. By definition, the compensated error amplifier gain must be equal and opposite so the net gain is 0 dB. I place the compensation zero one decade below Fco and teh pole one decade above Fco. Usually Fco falls in the 30 to 80 kHz range.
Let me know if this makes sense to you.