When the source is switched and unidirectional, the burden must also be switched and unidirectional.
An example application from my archives:
Source: my website, https://www.seventransistorlabs.com/Images/Mag_Amp_PSU.png
The bottom half of this can be ignored for present purposes.
Notice the CT primary (1T) is in series with the switching MOSFET, so the source is switched and unidirectional. The secondary (150T) has a large burden resistor across it (1k), which determines the peak reset voltage; then a diode into the nominal (small value) burden resistor.
Mind that this doesn't show phasing for some reason, which should be set in the obvious direction (transistor on = FR102 on).
As long as duty cycle is less than some maximum (the time required for the CT to reset into the 1k, several L/R time constants), the reading will remain zero-based.
Notice that another CT and diode could be wired-OR into the burden resistor. If you needed continuous current monitoring (say for an average current mode control), a second CT could be placed in series with the SBL2040, the secondary wired in exactly the same way (with its own reset resistor and diode). Pulse width would then have to be limited at the top and bottom end (i.e. it only reads correctly for, say, 5 to 95% PWM). I would generally recommend a Hall effect current sensor over two CTs, as it's less limited, but it is an option sometimes.