There is nothing wrong with those negatives:

So, where is that green fog coming from?

Here's an explanation: As you clearly shown on the third image film negative has an orange layer. (For an explanation of why that is read this link)

If you scan the negative and just invert the colors you are inverting that orange layer as well, and that will give the image that greeish fog.

Most folks would want to color correct and add magenta or do other kinds of distortions to the color of the image, but the key concept is that the orange layer is not part of the picture

In for a longer explanation:

On a negative you would expect the white areas of the image to be recorded as black, and therefore and the black areas should be white.

The orange layer is the unexposed part of the film, no light information has been recorded on it. It is the considered the minimum density of the negative. When turned into positive it should represent the darkest possible value: black.

So... shouldn't the area with the orange be white instead of orange?

It is that simple: that orange information has to be converted into white before we invert the picture, and to do that the information for the orange layer has to be discarded or at least turned into a neutral value that represents the brightest possible point in the image.

Using the curve tool (it can be also done with the levels tool too) find where the pixels for the orange layer are on the histogram, and move the max value of the curve to clip the information so that the max values are white.

Really? discard that much? Yes, is not part of the picture, it should be white or a very very very bright neutral color, so that when it is inverted it represents black.

Once you got rid of that extra information, invert the picture. Your black levels should be correct and neutral and the ratios between the colors should be ready to do proper color balancing.

Note that because of the relative low density of negative film and because you will be throwing away a lot of information, and stretching the curves (or levels) significantly, it only makes sense to digitize using a high bit depth of 16 bit per channel or better (AKA 48 bits per pixel or better)

The best place to find the values for the orange mask are in the edges of the pictures, but a segment of unexposed film can save you a lot of guessing.

When digitizing a full roll of film is surprisingly easy to get consistent results. Once you found the values for the orange layer, you can pretty much use that setting for the whole roll and it should work (For that I recommend scanning or digitizing with a fixed exposure and avoiding any kind of auto-color auto-exposure or auto-anything)