Ah, someone bites 8-) Really the only thing they can guarantee is that they're not intentionally bringing peanuts into their facility (ie they're not mfg peanut fillings or coatings). In many origins, it's quite common for a farmer, grader, middleman, dockworker, or ship-hand to eat nuts as a snack. The shells and meats of those nuts are often tossed aside, and i can't count how many times I've seen them land in piles of cocoa or in bags. In some cases the equipment used to handle the material is often shared between nuts and cocoa.
How does the mfr propose to keep that material - which they're not even aware of - out of their plants?
If they make the argument that it's too small an amount to be reactive - how do they know what the acute dosage threshold for an immuno-reaction is for each person? How do they know that even if they're testing it via ELISA or some other assay, that a given persons threshold for a reaction isn't below the threshold of the testing kit (ie absence of assay reaction doesn't guarantee it's not there, it just says it wasn't in that sample or that it wasn't enough for that particular kit to detect)?
I would be very skeptical of any non-artisinal mfr (who isn't in complete control of their supply chain) guaranteeing the absence of nuts in chocolate. I don't believe it can be done on a large scale, and would thoroughly challenge any mfr who claims they can do so.
Winnowing is not 100% effective. In fact, in some cases, it's pretty darn ineffective. You will always have some level of shell material in your nib. And your prewinnowed stream always will have some nibs in it. Meaning that some % of the nibs would come into direct contact with any foreign material (in this case, pnuts) - so removing the shell, even if it were 100% effective, wouldn't guarantee a soln for contamination.
Of the two companies you mention, one makes chocolate, one does not. We can throw out the company that doesn't make chocolate as they really have no feel for what's involved, and are making assessments based on the information they have (incomplete). It's a bit like me selling used cars and guaranteeing the previous owner(s) never took it on a trip longer than 30 miles - i have no idea. I can provide you a statement to that effect, but it doesn't mean it's true. The other company you mention doesn't really have the technical competencies to provide an accurate assessment (sounds harsh, i know, but most companies fall into this category. not meant to be harsh).
Regarding the challenge study - what size was the sample tested? How many studies were done? On multiple lots? Over extended periods of time? What was the test used and it's minimum detection level? Are you certain that because the test didn't register above the minimum detection level that it's not present at slightly below the detection level? When a given lot does test positive, what are the appropriate clean-out and flush procedures to guarantee 100% elimination of any residual proteins? Are you confident they even can be removed 100%? And now we're back to the threshold of detection level question - if post clean-out it's not detectible by the assay, does that guarantee it's completely gone, or just that it's below the detection lvl of the assay - but still high enough to kick off the IgE cascade?
I've spent some time on the allergen issue (not just nuts, but dairy and soy as well), and know the answers to the above questions. It is my professional opinion that it is simply impossible to guarantee - with any level of validity - the absence of nuts in chocolate. A written statement by a company to that effect is simply reckless.