Slightly overexposed

In both case, your camera did a good job in the exposure calculation. The camera doesn't have the dynamic range of a human eye and at some point it has to do a compromise which is either clip the dark background or overexpose the luminous flower. Since the camera use an evaluative metering, both background and the subject are suffering from the compromise. I think the best solution was to keep the camera in "evaluative metering" and use an exposure biais of -1/3. Then use some post-processing to "distort" the reality! Forkball attempt is very good by the way. My point is; the job of the camera is to capture as faithfully as possible what is put in front of the lens!

Take a look at this picture (no retouch, only resizing):
http://img144.imagesha​ck.us …/7794/mtlnoreto​uch2ex.jpg

Ok its not interesting but the point is that it looks very wash out. Colors are flats, no contrast, no saturation, etc... But this is what I was looking at. The sky was kind of white with a lot of humidity in the air and I have to admit that this shot is really accurate.

Now, take this same shot with modification to the "levels" and "saturations".
http://img144.imagesha​ck.us …938/mtlwithreto​uch4io.jpg

Wow, it looks like a Nikon shot!! The picture is more pleasant to the eyes but I still believe this is not the job of the camera to give you this kind of picture unless you play with the camera parameters to accentuate contrast and colors saturation.

You just solved your own problem. Spot metering on the flower in this case would have shown you were probably right on or maybe even 1/3 overexposed depending where on the flower you put the spot. Keep shooting.

I would have thought partial would be ideal for the flower shot, the circle in the middle would represent pretty much the full range of the tones on the subject? But then I suppose the flower would then end up 18% grey, which may not be correct. I think I need that book......

Your "overexposures" are the result of the camera trying to lighten up the vegetation in the backgrtound. In exposing for that, the flower in the foreground gets more exposure too. You might find spot metering will go too far the other way, especially with a light flower like in your first example, where spot metering will be trying to get the tonal values of the flower to about what Evaluative tried to get the leaves to.

So part of the photographic problem here is something the camera is just too dumb to understand. As far as it's concerned, everything you try to take a picture of is a neutral grey. If that's not the case, you have to tell it (by dialing in exposure compensation) that it's lighter, or darker, than "normal". But then if it were easy, everyone would be able to do it. So you were doing the right thing in your two examples. You just didn't know why, but you instinctively knew what to do. And that's very good. Too many of us realize what we should have done only when we get back to the computer.