Curious - it's the alpha channel (the mask) that you're seeing in the second image. I've just repeated this with Affinity Photo, and also with several of the other stock G3 characters. (I'm currently using CTA3.02.)
You'll find there are two images opened in your image editor. #1 is the alpha channel data and that window is titled foo_Opacity.png. That's the one that primarily black and white. #2 is the main image window and is named foo_Diffuse.png. Unlike your experience, I found that editing foo_Diffuse.png and saving it (a simple cmd-S) saved the changes straight back to CTA3.
Some of the characters are "cleaner" than others. Try Elastic Folks Female and look at the split images in Photoshop/Affinity Designer/etc for a nice clean example. Then once you can relate the mask image, take another look at your Ted character and see how the mask is doing more work to produce the final image. At the furthest extreme, take a look at the Caveman components - in this example, it's the mask doing almost all the work to produce the final image.
I know that with time that has passed since you posted your original enquiry that you have possibly sussed this and moved on. But with new users coming on board, I thought it might be worthwhile 'documenting' this for their benefit.
CrazyTalk Animator 3 Pipeline (always the current production version, unless otherwise stated)
macOS High Sierra (10.13.6)
Signature data last updated 2018-10-06