In the notes from your data sheet on page 2 it says,
2. Inductance shown for each winding, measured at 100 kHz, 0.1 Vrms, 0 Adc on an Agilent/HP 4284A LCR meter or equivalent. When leads are connected in parallel, inductance is the same value. When leads are connected in series, inductance is four times the value.
You would expect this because L = Al.N^2, where Al is the specific inductance per root turn. So you double the number of turns you get four times the inductance.
I don't think you will find that the coupled 100uH inductor behaves or rather performs in the same way as two separate 200uH inductors. Looking at the data sheet circuit for SEPIC,
**broken link removed**
In DC terms C1 must charge to VIN. With Q1 on then VIN is placed across L1 and also the same voltage is forced across L2 by C1, given it is charged to VIN. Visually it looks like L2 has -VIN across it but the 'dots' are such that effectively the pair of inductors are being driven with the 'same' voltage in parallel.
You do not want 100% coupling in this application since it would force currents in the inductors to be discontinuous saw tooths. It is the leakage inductance that prevents this from happening and also allows you to take advantage of 'ripple current steering'.
1980 - 2003 Unitrode Design Seminars
**broken link removed**
SLUP103
SLUP104
SLUP105
Of course the devices you are looking at will not be for 200W PFC applications but the principles remain the same. If you do wish to implement ripple current steering then as per the above papers given the SEPIC output is noisy anyway you would wish to steer ripple current into L2 which means the leakage inductance would be placed in series with L1.
Depending on how the devices you are looking at are manufactured then that leakage might be equally distributed. Otherwise you might wish to measure each winding with the other shorted to determine where it is located. The short circuited winding with the highest measured inductance referenced to the other would be the one to place at the input in your design.
Having determined what that 'leakage' inductance is you will be able to work out what sort of damping is required across C1.
Genome