Another issue is WS procs. It is believed that WS procs are actually at a bit lower rate than JA procs, even though it is usually possible to use JAs much more often than WSes.
From my parses on pup, weaponskill proc rates are also at about 20%. (I got 18.8% overall, but that includes the puppet using its abilities on proc'd mobs; it excludes pure-damage weaponskills on proc'd mobs, since I use different weaponskills for procs vs damage.)

The main difference is that weaponskill procs (aside from puppets) inherently do damage to the mob (plus counting the melee time needed to get TP for another attempt), which means the mob will always die faster from weaponskill procs than JA procs, reducing the chances of getting a proc before the mob dies. Number of attempts per mob determines the overall proc rate, so that's an important factor.

I've never parsed magic proc rates, but from incidental attempts that I've seen... Well, ninjutsu seems to have a very low proc rate; 10% at best. Normal magic from rdm (when silencing/sleeping/etc mage beastmen) doesn't seem too terribly bad, though still annoying; dia almost never proc's, but things like silence/sleep/paralyze/slow/etc seemed more likely.

Personally, I'd probably suggest having the chance of proc'ing be proportional to the cast time of the spell (base cast time of the spell, not the result after fast cast is applied). Things like Dia and extremely fast blu spells would be on the low end. Maybe ancient magic could be like a 50% proc rate or something. And maybe it already works that way; I don't know that anyone's done any testing on it other than the gut feeling of the above anecdotes, with obvious eyeballing caveat. It might also be proportional to damage done, for spells that do damage (and then a fixed value for enfeebles), which would also make fast, low-damage spells like Dia less useful.