It's possible that the disparity is due to viewing different time frames. I was looking at around 1.5mS, where there are still surge influences. Things settle down under static conditions if you wait for
15mS. If loads were only static....and the simulation was remotely accurate.....
A pencil and paper are quicker, and more portable. Was off topic here, originally thrown off by a reference to measuring 6 turns on LV primary - you meant 3 turns, which made more sense, but by the time I figured that out....
You know that Lp and Ls can swing simply due to micro-gap quality / cleanliness of core butt joints etc... Leakage inductance doesn't. If you use the coupling coefficient to set leakage inductance, as the simulator does, it WILL swing just as wildly as the Lp being used in the simulation.
The relationship used by the simulation is
Llk= (1-k)Lm , where
Llk is the leakage inductance seen in one winding when the other winding is shorted. Lm is the magnetizing inductance of same single winding. k is the coupling coefficient.
Note that leakage inductance effects in the secondary use Lm and Lk measured on the secondary winding, with the primary being opened or shorted, as required. While the two leakage terms are normally related by the usual N^2 factor, low voltage structures' lead-out effects can be signifigant, as they form part of the turn structure and are often missing from the measurement.
One last thing, in a full wave rectifier fed by near 100% duty, stress on the filter inductor is usually pretty small. Even so, in this simulation the current zeros during the phase change. This means you should probably check for core loss in the filter choke, before sizing it.
RL