TL;DR: the dimmers aren't switching off completely: they're allowing some current to leak through, which is why you're seeing a voltage across the CFL. A different make of bulb may behave better with the leakage current that you're getting. Or perhaps a different brand of fan (if you haven't installed them all already).
I do know that operating CFLs in those sort of conditions will shorten their lives considerably, so you might actually be cheaper for you to use incandescents instead (a quick calculation says about 12 kWh per year for a 60 W bulb).
Read on for the technical explanation...
This is a circuit diagram of the innards of your fans:
The voltage across the bulbs, Vb is determined by the formula:
Vb = Vin * Rbulb / (Rdimmer + Rbulb)
where:
- Vin is the mains voltage (120Vac or 240Vac depending on country).
- Rbulb is the resistance across the bulb or bulbs.
- Rdimmer is the resistance across the dimmer.
The dimmer is a solid-state electronic circuit, so it has a very high effective resistance -- 10s of megohms is not unreasonable. Ditto for the control circuitry in the CFL. An incandescent bulb is a simple piece of resistive wire; a 60 W / 120 V bulb will have a resistance of 240 ohms.
Now, suppose the dimmer has a resistance of 50 MOhms and the CFL has a resistance of 10 MOhms; plugging the numbers into the equation above gives you 20 V across the bulb. OTOH, the voltage across a 60 W incandescent bulb will be about 600 microVolts, nowhere near enough to make the bulb glow.
If you have two bulbs in the light fixture, the resistance, R, of the two in parallel is given by:
R = R1*R2/(R1 + R2)
So if you have a CFL and an incandescent installed, the effective resistance is going to be very close to that of the incandescent alone:
R = 10,000,000 * 240 / (10,000,000 + 240) = 239.99 Ohms
Again, not enough to turn on either bulb.
With two incandescent bulbs, the effective resistance is half that of a single incandescent, so you have half the voltage across them.
The flickering you see with two CFLs is because the light you see is basically a high-voltage spark through the tube. The CFL contains circuitry to amplify the incoming voltage up to the point where the spark can occur. Under normal circumstances, the input voltage is enough to cause this spark 100 or 120 times per second (depending on mains frequency), which is far too frequent for the human eye to notice. With the reduced input voltage, it takes longer to reach the required voltage, so you notice the flicker. No two bulbs will be exactly identical, so they'll flicker at different rates and take different times to recover between discharges.
The basic problem is that the occupancy switch that you chose to purchase is one designed to be able to work in series with the load attached to it. If the attached load is too small (too low of wattage) the switch assembly does not get enough power to operate correctly. For the particular model you have this is on the order of 40 to 50 watts of minimum load. From a technical standpoint it kind of sucks that they work this way BUT they are designed to work in place of a simple light switch that is in in series with its load.
Your Radio Shack relay does not come close to representing a 40-50W load on the switch so the only other way is to add some additional load in series with the switch (i.e. equivalently in parallel with the relay coil). As you have come to see this is not a very satisfactory solution and some other scheme needs to be found.
The best situation in this case is to acquire the alternate type of occupancy switch as mentioned in that thread that you linked. That type of switch has a live AC lead and neutral AC lead connected to it to power the switch unit irrespective of the size of the load. A separate output connection goes off to the load you are trying to switch. This type of switch may take a little re-wiring in your garage but that should be easier than trying to wire in the kludged extra load that the current switch requires. If you choose to deploy this improved version of the switch you will be able to remove the Radio Shack relay from the installation.
To make yourself feel better about the fact that you now have an extra occupancy switch you could pick a room in your house that has a tendency to get its light left on accidentally and install the extra unit in place of its light switch. Maybe this is a bathroom the kids use or a far closet that does not get used often.
BTW
It looks like Leviton now has a whole slew of different model occupancy switches that are also designed for single pole switch replacement but have removed the minimum load requirements. These do not require a neutral wire connection but do require a GND wire connection.
Best Answer
I once performed a direct study on this subject. I built an apparatus to power-cycle a T12 fluorescent lamp every 60 seconds and an identical lamp every 12 hours. I used both recording watt-meters and standard household electric meters to measure the power consumed.
I expected that the power consumption of the filaments during preheat would be different from the power consumption of the ionized arc. Probably more, but possibly less. I was surprised to find that even after 96 hours I detected no measurable difference.
I could only guess that the ballast limited the filament current just as it limited the ion arc current.
This was so long ago that there was only one kind of ballast available, what we now call a dumb ballast. I don't know what the result would be with a modern fixture.
I also applied an oscilloscope to the line side of the ballast and I could see a power spike that lasted less than 1/120 second. I couldn't get a good idea of its magnitude. I theorized that the meters used moving parts and could not practically respond to an event of such short duration.
My conclusion was: Yes, a fluorescent lamp does consume a power surge on startup -- but you don't pay for it!