In my experience there is often an assessment process that is essentially just made up. And when properties do sell, the sale price is always a "surprise" relative to the assessed combined value of the property.
In a sense the question is "What in particular makes you confident that the estimate accurately reflects the price of the land" but in a deeper sense what does the concept of "price of the land" even mean in practical terms? How would you know that the answer is right even if you were omniscient? And given the practical divergence from whatever theoretical standpoint, does then this value serve the same objectives as a "true" LVT?
So is your argument that you don't understand something and so it must be wrong?
There's an extremely large amount of existing material that is used by property assessors available for you to look up to research how they do this. It is a well-established field.
Property assessments have been done across a huge number of countries for decades probably billions of times at this point. There are probably trillions of dollars of capital that flows according to these assessments.
At least for an entire value tax, we have the benchmark of "when a house is sold, how closely does the price hew to the imputed value for tax purposes" to determine if the tax is, in some sense, fair. Some municipalities do better than others by this benchmark, but many of them continue their totally broken methodologies (by the standard of predicting sale prices) for a variety of political and bureaucratic reasons.
For a specific attempted implementation of a land value tax, how do you go about measuring whether you are doing it well? You can sometimes get sparse data points when vacant lots go up for sale, but otherwise you're just benchmarking against other models.