Take your transformer......draw a pin diagram of it.....number each pin, then measure inductance between the pins and draw in the relevant coils connected to the pins.
Then when you have done this, you arbitrarily put a dot (for phasing), on any pin that you like. I will then tell you how to find out where the other dots go on the other coils.
....but i give a clue....you connect two coils in series....the way of connection which gives the highest inductance is the way that means those coils are connected dot_to_no-dot.
You understand?.....its all about whether or not the fields from the coils in the core cancel or not.
Unless of course you have a datasheet which gives the dot locations for you?
That lightbulb test you do is bad, please do not do this.
- - - Updated - - -
and no you dont need to connect the full input filter, but you do really need to have some decent amount of input capacitance very tight up to your power switch loop.
And of course, pcb layout issues.....you dont want switching power current running through lengths of control ground.
You can of course, sometimes get some mileage out of a bad layout by increasing the fet gate series resistor....but not so much that you overheat the fet due to switrching losses.
I must admit id prefer a rcd clamp than what you have for the primary, just an r and c .