This page was inspired by the discussion on the (now-defunct) Canon EOS mailing list in November 1996 regarding whether the EOS flash exposure actually depends on the location of the subject in the frame. The earlier discussion was limited to E-TTL and FEL.
I performed some observations in late November of 1996, and here are the results. The short answer is: yes, I could observe the difference.
Method: Removed film from camera. Used both the built-in flash and a hot-shoe mounted Sunpak 433AF (which works in TTL mode). A "scene" was composed of a white pillow about 1 meter from the camera, filling about 1/3 of the frame. The rest of the "scene" was the back wall about 5 meters away. Tv exposure mode was used with a shutter speed short enough to make the aperture value blink, indicating insufficient light.
Since I have no flash meter, flash exposure was estimated by measuring the flash recharge time after exposure. With the built-in flash, differences in the audible recharge sound were evaluated.
When manually selecting the focus point on the white pillow, the built-in flash recharge sound could not be heard. The external flash recharged in 2.3 seconds (average of 3 measurements).
When manually selecting the focus point on the back wall, the built-in flash recharged very audibly. The external flash recharged in 6.7 seconds (average of 3 measurements).
Similar results were obtained by letting the camera select the focus point. By moving a high-contrast object on the scene, the focus could be obtained on either side of the frame, and recharge times similar to those above were obtained.
Generalizations: with plain TTL, the system appears to work as claimed by Canon. I have no means to check E-TTL or A-TTL.
Previous tip: Are EOS apertures stepless?