Yes, through M2's body diode. Then M2 comes on and shorts C3 to the primary's flyback lead, during flyback.
The switcher's outputting lots of runt pulses. That's not happy, and it's hurting your efficiency.
There seemed, on visiting Digi-Key, plenty of choices for suitable low-voltage drive FETs.
Is my download right, C3=10uF? Trying to directly connect the flying-back side of the primary to 10uF, during flyback, right?
That's obviously not right. You're dissipating the flyback pulse, which you shouldn't. The snubber only has to dissipate the energy stored in the leakage inductance. If you're dumping a lot more than that, you shouldn't be.
The leakage L for your transformer is 40uH (max), which is only 3% of
1,310uH. K = 0.83 in your sim, far too low.My 'crude' version was dissipating 1.2W, IIRC. You can drop R7--that was just for simulation convenience, to damp the ringing. R10 then dissipates 180mW @ full load.
Compared to the active snubber, the passive version saves two ICs, an auxiliary power supply, and a power FET. Seemed worth considering.
Don't forget the LT3757A's UVLO--you'll want to use that, too.
Cheers, James Arthur