One of the shortcomings of your test is that the smoke and combustion products of the incense is greater in volume than the air entering the system.
You are actually pressurizing the interior of your model, albeit a small positive pressure.
Therefore, you will see the smoke issuing from all the upper openings - not unlike a chimney.
If you want to conduct this experiment more accurately, you should run the incense outside of the model and allow the smoke to be drawn into eaves and gable vents.
Another shortcoming is that in fluid mechanics modeling, the path of the fluid must be scaled differently than the rest of the model in order to make the Reynolds number for flow similar in the small model to the full size case. This principle is called “similitude” (see http://en.wikipedia.org/wiki/Similitude for more info) and without it a scale model will not represent the operation of the full size case.
However, all of this is moot because you cannot see the true measure of the effectiveness of a ventilation system.
That measure is simply - how well does it cool the roofing material.
No matter how much air flows into and out of your attic, it’s not cooling your roof if it isn’t flowing against the roof.
This brings me to the final shortcoming of this experiment. There is no wind and therefore no pressure differential between one gable end and the other. Even imperceptible breezes create measurable pressure differences between openings on a house. Air flows from regions of high pressure to regions of lower pressure and it will follow the path of least resistance. Thus, the short circuit.
So to clear things up, gable vents and power vents will provide more airflow in terms of volume than soffit/ridge vent. The problem is that this air does not flow uniformly under the roof and therefore, can not cool the roof as well.
Air cools by convection and to have convection you have to have contact with something at a different temperature. More contact, more convection, more cooling.
If you’re looking for scientific evidence of the short circuiting look no further than the principles of fluid mechanics and thermodynamics.