Brian, on the one hand, you seem to agree to the "trade-off", by acknowledging that Fisher chooses to lump those iffy TIDs into the "good" category "just to be safe", as opposed to Whites (XLT for instance), that chooses to put those iffy TIDs into the reject category. You acknowledge that this will, of course, mean that the Fisher user would therefore be fooled more often (albeit "safe").
Yet your opening line is: "I don't think I would agree that the discrimination is weak." Seems to me, that if a person can't really discern deep nails vs deep coins (aside from some risky tedious sound/pinpoint tricks), then YES, that to me, would be defined as "weak discrimination". I don't know how else to define it.
BTW, using the XLT as an example (or using an Explorer, and simply lowering you iron mask way low into the potential false signal category, etc...) you can open up more #s down into the iron range, and open up +95, etc... and accomplish about the same thing. Ie.: some nails, depending on the way their bent, how deep they are, moisture of the soil, size of rusty clump, etc... you can indeed get a signal off other discriminators, from rusty iron. Just depends on your setting. And XLT, Exp users, etc.... do indeed chase a lot of those cross-over signals, when they are in a relic mindset, chasing super deepies, etc... But on the other hand, with the CZs, where those fringe iffies are just automatically put into those categories, seems you have much less user info. Ie.: either beep, or no beep, with no inbetween audio or cursor movements to make judgements from.