As you regard the answer as obscure, you're invited to calculate the exact transfer function of a second order filter build with a band limited (let's assume first order for convenience) amplifier by yourself. I hope, you'll immediately believe without calculation, that the transfer function must be of third order then.
As the most interesting question, in which way and how severe will the ideal filter transfer function be affected by the finite GBW? To anticipate part of the result, the filter cutoff respectively center frequency as well as it's Q are changed. The said factor 50 to 100 is a rule of thumb if you want to keep the ideal filter parameters with acceptable tolerance.
If the factor isn't feasible, you can adjust the filter dimensioning to consider the real amplifier. Because the real transfer function is actually third order, you possibly can't fit exactly the intended characteristics.