Without being able to see the negatives in question, there are three things you should check:

1. Monitor calibration: This can be ruled out somewhat if you can make two identical scans with the same parameters (curve, white/mid/black points, etc.).

2. Backlight on your scanner: I have had issues with my own scanner where sometimes the backlight is set very bright and shows up in scanned slide images as overexposure in the blue RGB channel. In a negative color scan, this will manifest itself as a very bright or overexposed image as blue varies with the neutral density against negative color film. Use the histogram in silverfast to see where your samples actually lie in the resulting image versus your scan parameters before scanning. I stopped using my scanner once it became evident that this issue was random even when the histogram showed what should be a well-exposed or ETR image and the scanner couldn't deliver.

3. Color Profile: Check to see that you're not editing images that are in a custom color profile tied to your scanner. For your needs, you will probably want the output profile to be Adobe RGB or sRGB. This issue is likely to manifest itself by encountering dull underexposed images or images with areas that have significant unexpected sample clipping (255s and 0s in 8bit channels).

It is possible that your negatives are also damaged or altered in some way, but you can check this by making and scanning test shots against a color target (such as an x-rite). Portra is considered to be among the more neutral color profile films, so if you are seeing too many swings in exposure rendition, you are probably having issues between the scanner and your final image, likely silverfast or the scanner driver is making assumptions in applying gain from the sample data, or you have a bad scanner as mentioned above.

Regarding your fifth point, it may be possible to get pretty close to 100% neutral rendition in image midtones with Portra (particularly Portra 160, the least saturated of the Portra family), but most film emulsions actually have less dynamic range and different color response and contrast than DSLR / digital images, which can get much closer to a realist interpretation. What you interpret as good portra 400 may actually be the result of people using similar workflows that are outside of a straight 50% gray metering and calibrated workflow. Some people even overexpose portra 400 for the flat highlights for wedding photography. Even Kodak has mentioned that the goal of the modern Portra line is to be relatively color-neutral and to let the saturation of the final result be determined by a workflow.

Thus, consider shooting a color test card to see the actual variance versus core colors in film emulsions for yourself, and also make sure to take test shots in varying types of light (outdoor x daylight/evening/night, indoor x florescent/incandescent/flash). Without physical filtration, film does not have any natural white-balance built in (except for special tungsten-light editions, which are not made anymore). You can edit 50% gray in most images to get to the point zero you mention, but not the entire color gamut rendered by the film as every emulsion is designed with its own frequency response to the primary wavelengths of light. This is easy to observe when switching to slide films or orthochromatic vs panchromatic black&white films, and some are particularly far away from neutral, such as Fuji Velvia and anything developed using experimental emulsions (lomography) or development (cross-process, etc).

Footnote: One can't scan a gray card in a scanner to calibrate it because film is translucent and always gets scanned with a backlight in addition to the scanning light. Hence, you have to shoot something that is 18% reflectant / 50% sample gray with your medium, then develop/convert it and see how it is interpreted by your workflow.