There is NO direct relationship. It is like the difference between power and effective radiated power. If I have a 100 watt transmitter, it delivers 100 watts, if it goes into an antenna with 10db gain, it has an EFFECTIVE RADIATED POWER of 1000 watts. To someone looking at the antenna in main lobe, it looks like I am transmitting 1000 watts, to someone who is way off the main lobe, I may appear to be transmitting only milliwatts. Candlepower is the apparent brightness as if the light source was an isotropic radiator.
The optical or reflector assembly is anything BUT an isotropic radiator. The narrow the beam produced, the higher the ratio of candle power to lumens.
An intensity of 1 lumen per steradian(sr) is 1 candle power.
So a 12.6 (4 x pi) lumen isotropic radiator would produce 1 candle power. If I make the beam .001 SR (and the optical/reflector assemly is 100% efficient) , the 12.6 lumen source will have 1000 candle power. If I have done the math right, .001SR is about 3.5 degrees, so that would be a fairly narrow beam.
Candle power is roughly lumens /beam angle (in steradians if the reflector or optical assembly is reasonably efficient.. I.E. the ratio of candle power to lumens is the ratio of apparent brightness to real brightness. A sphere is 4 pi steradians.
So the short answer is there IS NO DIRECT RELATIONSHIP BETWEEN LUMENS AND CANDLE POWER.