DiMAGE A2 Exposure Reliability

My test results showed unexpected variations of image intensity that implied errors of exposure. My first reaction was to blame myself, and to be as meticulous as possible in metering exposures, assuming that the camera itself is precise and accurate. I am now questioning the camera; so far the results suggest that the camera's metering system may mis-expose by up to 1.5 stops.

Metering technique. My early tests were done by holding the camera close to a middle-gray target, setting the exposure dial on "A" to use aperture priority, press the shutter button halfway down several times while moving the camera slightly, and observing the shutter speeds that were called for. When I achieved consistently the same shutter speed with repeated measurements of the target, I moved the exposure dial to "M" (manual) and adjusted the f-stop and shutter speed to the values shown when it was set to "A".

When the results using this were inconsistent, I set the camera to use spot metering, and used the viewfinder to ensure that the camera was aimed at the middle-gray target, and took repeated readings as before.

The results:

I shot a series of images to see what the camera's contrast control did, and was embarrassed to discover that the images seemed to be accidentally underexposed by about 1.5 stops.
Here are a correctly exposed (left or first) and an incorrectly exposed (right or second) image of the same target:

The image on the right is about 1 stop under-exposed. This is concerning because I used the same part of the target to meter both exposures. If the meter is reliable and the technique is the same, the images should have the same density.
The following set of contrast curves relates to the image on the right above. Underexposure is revealed by the fact that the right half of the curve is steep and pixel saturation isn't quite reached, and the left half of the curve is long and shallow. The image intensity represented by the red dot on the Zone V line is taken from the upper left dark region of the target, the red dot on the Zone VI line is taken from the upper right lighter region of the target.

The camera's histogram at the moment of exposure indicated slight underexposure, as the intensity peak for both halves of the target were below the midpoint of the graph. But the photographer had a short time before taken careful readings, and the idea that the manner of taking the exposure reading may have been improper or that the light may have changed did not at that moment well up to the top of his consciousness. Even less was he prepared to think that the exposure algorithm favors underexposure.
Next, I shot the same series in cloudy-bright conditions and got slightly less under-exposure:

This image doesn't show pixel saturation until Exposure Zone IX, as desired, but the target values the midpoint yield rather dark, low-contrast image intensities. But this can be easily adjusted with the Curves Tool, as no values are lost.
These two series of exposures also compare "noise-reduction ON" with "noise-reduction OFF". It is wonderful to see that there's no increase of image intensity in the darkest values at this ISO -- no detectable noise.
When I plotted the curve from the lighter half of the board, I got a curve closer to what was expected -- middle target luminance values with middle image tones: .

But this set of curves is obtained by taking the exposure reading from a middle-gray target (the dark half of my target board) and taking image-brightness measurements from the image of the bright half of the target, which is almost exactly one stop brighter than the dark part of the board.
Then, I took a series in bright sunlight. This produced a series of images that are overexposed by about 1.5 stops!
First, compare a correctly exposed image (left or top) with an overexposed image (right or bottom) taken in sunlight that (based on the metering) was expected to be equivalent:

The upper left dark area of the target is represented in the characteristic curves below by the red dot in the Zone V vertical line.

Remember, the "Exposure Zone" is the placement intended by the exposure reading, not the resulting image intensity.

This result woke me up to the possibility that the camera's exposure algorithm or metering function may be unreliable, as I was particularly meticulous, setting the camera to use spot metering and carefully taking the reading from the dark side of my target.
Because my target board's bright half is almost precisely 1 stop brighter than its dark half, I measured the image intensities of the same exposure series and discovered them to be much less overexposed:

In fact, this is still a half-stop overexposed compared with the standard:

Next topic: Change in Contrast with custom White Balance