That's not the "most accurate", because incident metering makes assumptions about the reflectance range of the subject matter. Specular reflection (jewelry, mineral based makeup, rhinestones on clothing, dew on plants) renders those assumptions invalid.
Hi Joseph
Please help me to understand this... I thought that both reflective and incident metering make assumptions about the reflectance range of the subject matter; that they both 'reference' a standard 18% reflectance. Is that not true?
Incident makes no assumptions about scene reflection. By measuring the light falling on the subject, it will put mid tones in their proper places, no matter what percentage of the scene they occupy or what the scene's average reflectance is. A tiny little 18% or 35% reflective face will be properly exposed even if it's in a sea of white or one of black.
But incident will not account for either the dynamic range of the camera or the scene, so a 120% white like a wedding dress will blow out, or a 500% reflecting piece of jewelry.
TTL, in it's simplest and oldest form did make an assumption about the scene being 18%. If you have a Nikon D100 or D1X, that's basically what you got. Newer TTL systems work differently. The iTTL system uses whatever the camera has for light measurement, up to 1005 zones in cameras like D2X or D3. It should be able to recognize scene elements, like several small flesh colored areas are faces, and base the exposure on them, instead of trying to bring dominant scene elements to 18% gray.
I am a wedding photographer. When I use Nikon's iTTL (not-BL) on my SB800s and am shooting, as an example, a horizontal close up of three black tux clad men, the flash always overexposes by about a stop, as its trying to bring the reflectance of the scene to 18% grey, though at least two thirds of the frame is black tux cloth.
It shouldn't be. Did you set "center weighted" metering? Normally, it should also use "3D metering", checking focus distance to "sanity check" the flash power. The 3D feature disengages when you turn the flash or tilt it up or down (because you can't pre-calculate bounce flash) but if you have the flash pointed straight forward and put some sort of light modifier or diffuser ("mini" soft boxes are the worst) the 3D feature totally screws up exposure.
Conversely, when shooting a close up of a white clad bride, it underexposes for the same reason.
Again, on the newer cameras, that's what Nikon's iTTL is supposed to avoid.
Both examples assume indoor shooting, camera in manual, with negligible ambient light.
Yet, when I take an incident flash meter reading of the scene and set the SB800 to the appropriate manual setting, the exposure is correct no matter what the reflectivity of the subject! (assumes the camera-mounted SB800 maintains the same subject distance.)
Whats going on?
Either iTTL isn't doing it's job (and it sometimes doesn't) or you're using a light modifier or a metering setting (center weight or spot) that's keeping it from working.
Best I can say, if you're working in situations where you have the time for incident metering
and the access to the subject position, it's a good way to go. But you really need to debug the TTL for the more "action" oriented parts of the shoot.
--
Rahon Klavanian 1912-2008.
Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.
Ciao! Joseph
http://www.swissarmyfork.com 
