is there really a such a thing as higher ISO than base on a CCD or is it just a metadata setting and no gain is being done on the chip.
It really depends on the camera design. Some CCD cameras use a variable amplifier before the ADC, some simply shift the digital value. With a well designed 16 bits ADC, the second system is not necessarily worse than the first.
A CCD (and Cmos) pixel is just like a bucket that will fill up with photons till the bucket is full. The full value is typically about 50-60 thousands photons for the sizes used in MF sensors (about 6 µm). With 16 bits, the ADC can count between 0 photon, 1 photon, 2 photons, ... till 65535 photons. So it can count all possible values.
Doubling the analog gain means that for each photon, we count 2. So the ADC will count: 0, 2, 4, ... 65534, when we have 0, 1, 2, 131068 photons. But we cannot get 131068 photons, the bucket never holds that many, so we lose some possible values.
Doubling (or quadrupling) the analog gain makes sense if we only have a 12 or 14 bits ADC. A 12-bits ADC can only count up to 4095 values, far less than the 50 thousands a pixel can hold. So, when we have lots of light and the well may fill up, we count the photons by groups of 16 and everything is fine. When we know that we may only get about 16 thousands photons, because it is darker, we may raise the analog gain and count them by groups of 4. If it is very dark and we get at most 3-4 thousands photons, we may want to count them one by one.
That is how it works when there is no noise (except the noise inherent to the discrete nature of photons), of course. But noise does not fundamentally change the model, it just adds a small random number of photons to each bucket (typically up to 10-20).
