The discussion you're referring to, in the C4 documentation, refers to the process that a filter uses for calculating a matrix multiplication. This is actually just a description of what the filter does to the colors in the image when it gets applied.
In fact, what's happening under the hood is that the colorMatrix: method sets up a CIFilter called CIColorMatrix and applies this to a C4Image. Unfortunately the source code for the CIColorMatrix filter isn't provided by Apple. 
So, a longwinded answer to your question is: 
You can't access color components for pixels in a C4Image through the CIColorMatrix filter. But, the C4Image class has a property called CGImage (e.g. yourC4Image.CGImage) which you can use to get pixel data.
A good, simple technique can be found HERE
EDIT:
I got obsessed last night with this question, and added these two methods to the C4Image class:
Method for loading pixel data:
-(void)loadPixelData {
    NSUInteger width = CGImageGetWidth(self.CGImage);
    NSUInteger height = CGImageGetHeight(self.CGImage);
    CGColorSpaceRef colorSpace = CGColorSpaceCreateDeviceRGB();
    bytesPerPixel = 4;
    bytesPerRow = bytesPerPixel * width;
    rawData = malloc(height * bytesPerRow);
    NSUInteger bitsPerComponent = 8;
    CGContextRef context = CGBitmapContextCreate(rawData, width, height, bitsPerComponent, bytesPerRow, colorSpace, kCGImageAlphaPremultipliedLast | kCGBitmapByteOrder32Big);
    CGColorSpaceRelease(colorSpace);
    CGContextDrawImage(context, CGRectMake(0, 0, width, height), self.CGImage);
    CGContextRelease(context);
}
And a method for accessing pixel color:
-(UIColor *)colorAt:(CGPoint)point {
    if(rawData == nil) {
        [self loadPixelData];
    }
    NSUInteger byteIndex = bytesPerPixel * point.x + bytesPerRow * point.y;
    CGFloat r, g, b, a;
    r = rawData[byteIndex];
    g = rawData[byteIndex + 1];
    b = rawData[byteIndex + 2];
    a = rawData[byteIndex + 3];
    return [UIColor colorWithRed:RGBToFloat(r) green:RGBToFloat(g) blue:RGBToFloat(b) alpha:RGBToFloat(a)];
}
That's how I would apply the techniques from the other post I mentioned.