I would think that “lost arts” can be broken down into 5 groups:
-
We know a number of ways that the ancients could have done what they did, we just don’t know which specific method they actually used. For example, exactly how trireme rowers were arranged - there are several workable arrangements, but we don’t know for sure which arrangement the ancient Greeks used.
-
We know how it was done in a general way, but are no longer willing to devote a lifetime to learning the detailed techniques, and also no longer have an existing base of experienced artisans who could shorten the learning period through training. For example, medieval swordfighting - the medieval warrior was good at it because he spent years learning from experienced specialists, and then went out and practiced in real battles.
-
The “how could primitive Egyptians/Mayans/etc. have done that without modern equipment” lost arts. This is mostly failure to realize how much you can do with a large amount of workers, a lot of time, and simple techniques. A good example is the Iron Bridge in England, the first large metal bridge, where a BBC TV team showed that the perplexing question of how the large metal support beams were put into place without using complex machinery was easily answered through use of a simple wooden derrick, a lot of rope, and a large number of men working in carefully controlled unison.
-
The “legend” lost arts, where the lost art never really existed (i.e. the huge designs marked out in flat areas of South America, which are only visible from the air - does this mean that the makers could fly? No, only that they knew some simple techniques for design layout. After all, if the designs were made for religious reasons as generally agreed, the makers didn’t need to see them as long as the gods could).
-
And finally, there are almost certainly some real lost arts. It is likely, however, that all or most are in the categories of detail techniques (the exact shade of an ancient glaze) or obsolete techniques (we have developed an improved method of achieving the same result, so no longer need to know the previous method).