I know what you mean, and I also agree with barb...

The problem is that computers can't intelegently "interpret" like a person can, only apply math formulas to a situation.

And Wilcom has good ones (no doubt) but the objects in the original EMB are basically the result of the puncher intelegently interpreting a design and deciding the best way to break up a design.
And each object is like a realization on behalf of the puncher, that this particular part of the overall design is best redered as an individual and particualar type of element ETC....

But when you load in a machine format design as all this object data has been discarded the software as no intelegent way of knowing where the original objects started and ended as all it is left with is a whopping big sequence of individual needle penetrations and the odd function ETC.....

So what you are essentially saying is that "wouldn't it be great if the software could somehow regain all this lost object data PERFECTLY and re-construct the original emb from this stitch list PERFECTLY." In essence see the design in the same way that an experienced puncher would.

That is a very tall order, considering software's current inherent limitations....
Perhaps when neural network computing takes of in 50 or 100 years it wont be so difficult... but we are not at the star trek stage yet.......

On a slightly different tack.....

I have wondered however weather it would be possible (weather or not there is any scope in the various machine file specs for various machine formats) to somehow include "pseudo meta tags" in the stitch data (which could tell the program where objects start and end.
Perhaps by placing an empty stitch between objects or any other creative way using all the functions available to achieve this with-out compromising its fuctionality ETC.....
So that machine formats out-putted by wilcom could be much better "reverse engineered" when loaded back in.

Theres my technical bend on your observation there.........