There’s a very good reason. Sales of buildings plus land happen all the time, whereas sales of just land are virtually non-existent in a lot of the places with the most valuable property (e.g. Manhattan). Without real sales to tether the analysis you are just guessing.
b) You may not be confident in your ability to distinguish the value of the land from the value of the structure, but your insurer is. If you want to know e.g. how much your home's value would improve if you finished out the basement or built a garage, they will go through it with you on the phone. Sure, it's something of a guess, but a very educated one.
Also if you know a building would cost $X to replace, but the plot with it sells for $X+$Y, you have some real information, not just guesses!
Suppose there is a plot of land which (from some omniscient point of view) would offer maximal benefit to society if a four-plex were developed on it. The point of a LVT is that, under a LVT, a four-plex would be developed on it, not because anyone forces it but just because that's what's in the landowner's best interest; whereas under the current system (in which the structure is also taxed) the landowner might find it more profitable to build a duplex instead. The deadweight loss is the loss to society when the landowner builds something that's in their best interests but not society's.
That's hard to prove for any specific piece of land, since we don't know what the omniscient-view "right" use of the land is. But if you accept the idea that there is an optimal use, and you you trust the whole idea of free market price discovery, it's reasonable.