In the example I provided, both "identical" sites are worth $500,000. Once improved, the land value does not change. However, a view adjustment may still be warranted when both sites have identical characteristics. This is because of how the improvements are constructed.
Under a different scenario, when comparing sites with dissimilar characteristics, the land value might entirely account for the view. Assume for example, two identical 2-story homes were built instead, but one of the lots was worth $450,000 because it is slightly smaller. There is no view adjustment, and maybe the market values of the improved properties are $1,500,000 and $1,450,000, respectively. However that may also not always be the case. Maybe they're both worth $1,500,000 and whatever lot premium there was for size goes away. Or maybe the lot premium is magnified and the improved values are now $1,500,000 and $1,400,000. The point is that market participants for improved parcels can and do react differently than those for vacant land.
I am simply arguing that if you estimate the full value of the land, then an additional adjustment for view risks double-dipping.
One might risk that, or one might risk under-adjusting too. You can't always look only to the land for all the view adjustments. And if you do, that is where the risk comes in.