I agree. There are must be some losses in translation, cos pictures are just approximations to the elements of thinking process. Pictures have some "cross-platform" features, but their expression power much lower than language's. Ex. consider that language 1 has expresson power x, pictures have expression power y (x > y). And language 2 has expression power z (x = z). Then we're trying to reduce x to y and then take it to z again.
Instead of this we must take x to w (w > x) and then reduce w to z.
A human translator going exactly this way.
Instead of this we must take x to w (w > x) and then reduce w to z. A human translator going exactly this way.