Electronic – Reuse LEDs from a LED bulb

The LED power may be limited by thermal considerations. Both the power supply and the LED are prone to failure. The former because of possible shoddy transient protection and inevitable death of filter capacitors (exacerbated by the typical high temperatures). LED life will also be greatly foreshortened by high junction temperature.

These things are a trade-off. My suspicion is that the LED rating of 3W you mention is not really achievable (with any reasonable lifetime) in the environment of the light bulb, and everything (driver and LED) is really close to the line. Here's a snippet from a Chinese "3W" LED datasheet:

The Power Dissipation is absolute maximum 3W at \$T_A = 25°C \$. It will have to be derated above 25°C at the bulb, so I think you can see that calling that thing a "3W" LED is somewhere between irrationally optimistic and an outright lie. While the bulb maker might be perfectly happy to run it right at 3W, if they die so fast that they are getting 40' container loads of lamps dumped back at them, they'll not be so happy.

The only thing I really expect to be true of the bulb is the (initial) lumen output (you can expect them to fade over time). If the lumen output falls short, especially out of the box, that is false advertising and I would return them. So far, from reputable makers such as GE, I have not been disappointed. I can't directly measure lumens, but I use lux meter to check light levels on work surfaces and so on, and can reasonably well compare halogen and LED bulbs with similar angles.

Perhaps the dimmer circuit won't work without a resistive load. Some thyristor dimmer circuits might not see enough holding current from a rectifier-capacitor load to stay on for the half cycle. You may be able to wire a fat resistor in parallel with the LED lamps and have it work, however it may require some experimentation to determine an appropriate value.

Here's a typical small triac datasheet. The holding current is specified as maximum 10mA or 20mA, depending on type. If a phase-control trigger circuit provides a short pulse, the MT2 current must reach the holding current during the trigger pulse or the triac will not stay on for the remainder of the cycle. (bottom line below).

Here (from Wikipedia) is what the output of a typical phase control looks like. The gate trigger pulse would take where the vertical red chevrons are located. (a) is the power waveform. (b) is the output with triggering near the beginning of the cycle (bright) and (c) is the output with triggering near the end of the cycle (dim).

I would try about a 150 ohm 2W resistor -- if you get a few you can parallel several and go up in ~1W increments. You may find that too high a resistance causes half-wave triggering (intermediate brightness), since triacs have different holding currents in the positive and negative directions.