The LTspice model for the LTC3723 doesn't produce driver outputs above 390kHz.
Even starting with the macromodel's test fixture and, and changing only C1 to 40pF-80pF, reproduces the same aberrant behavior as I encountered in my attempts to model my design. Following is the actual LTspice design file I am using. Please note that the feedback path was disconnected to coax the highest duty cycle the LTC3723 could produce at 1MHz.
