That's the thing though: The progress from 0 to 6 is not reflected in any score.
The theory of NPS is that what matters is what people say about you. If people from 0-6 are all going to say negative things when asked, then lumping them in the same bucket is reasonable. It may not be as good as a more subtle scale, but it may be much better than thinking the numbers are linear.
It's possible that one could come up with a mapping that's even better, of course. But NPS is simple enough that even executives understand it. A marginally more-accurate number that nobody understands is probably worse, because people will trust it less. The point of the NPS score is not theoretical accuracy, it's motivating change.
Even if it did, though, it's not clear to me that there's much difference in the utility of the NPS metric. Are companies with a lot of zeroes also companies that are sincerely seeking to improve? Would a more complex scoring system motivate more change? If so, does the benefit gained outweigh the extent to which the added complexity harms NPS adoption elsewhere?
In practice, if some company had an unusually high number of zeros relative to sixes and were very serious about change and the metric didn't shift much when a bunch of people moved from zero to six, you can bet that someone would explain this in a meeting and everybody would still be excited. So although I get that this would be a problem if NPS were the only number used, I'm just not persuaded that some sort of NPS++ metric would be any better in actual use.