These 12% is interesting but how does it explains
consumer grade cameras expose differently from higher end ones?
In metering articles, the latest Nikon use 3D matrix III metering which
looks at your scene and compares it to 30,000 images database to guess the best exposures
Meanwhile all Canon beside the 1Dx use the dual 63 layer zones.
The 1Dx uses what I think is similar to Nikon that is having a lot of sensors (100,000 rgb sensors)
and run them through some algorithms to guess the exposure.
One thing disappoints me is that Nikon lets you influence the metering system using AE micro adjustment
from the D7000 when Canon only lets you do so in the 1D series
Even the most basic matrix metering system uses a fuzzy logic to guestimate what you are trying to shoot, it is the main reason die-hards (and slide film shooters) had such issues with matrix metering when it came out, you really don't have a clue what the camera is actually doing. The other metering modes are predictable because they don't try to work out what the scene "is" they just use illumination values, either as a center weighted average, a scene average or a spot average.
So as a specific answer to why cameras appear to meter differently, if they do, I would suggest this answer;
- People that buy lower model cameras don't necessarily know what they are doing, factory default is matrix metering (semi intelligent) and it works to get a "correct" exposure because of that intelligence. High end cameras assume a higher level of understanding and exposure knowledge and the weighting of the "intelligent" part is less as the operator is expected to have a better grasp of exposure and the amount of compensation they actually want.
- Alternatively, matrix metering nails it (and it does most of the time apart from exceptional lighting situations) so is plenty good enough for everybody, but, there was such a backlash from some pros when it first came out the "intelligent" part was eased back in there cameras metering algorithms because pros are troglodytes and too slow to embrace "new" metering systems.
Maybe try exposing a Kodak card, which is known to be 18%, with a 1 series and a Rebel in matrix metering and then in evaluative metering, if my thought is correct and the fuzzy logic has a different weighting in the two cameras metering algorithm then the two matrix metered shots will be different but the two evaluative metered exposures will be the same. But that is a complete guess and I don't have Rebel. (Unless Matrix Metering is "smart" enough to know you are shooting a gray card! Which it could be, see that is the problem with Matrix Metering...........)