I found the book in downloadable form at http://www.uni-leipzig.de/~logik/gottwald/treatise.pdf , and I quickly glanced through the material up to and including the page you mention. The first observation is that the axiom you quoted as $AX_{RT}5$ appears in this version of the book as $Ax_{RT}6$, so, unless there is a typo in your question, an axiom has been added after the version you saw and before the version I saw (or deleted after my version and before yours). My version includes an axiom, called $Ax_{RT}5$, that, in effect, limits the truth values to the intended set $\{0,\frac1{m-1},\dots\frac{m-2}{m-1},1\}$. It's not expressed using a disjunction as in your question but rather using an iterated implication, saying that all the implications $J_v(A)\to B$, for all $m$ truth values $v$, together imply $B$. So, since you seemed concerned about the absence of such a limitation on truth values, I would guess that this axiom might have been missing in an early version of the book and then added later.

You were also concerned about Gottwald's use of the axioms $J_s(\bf s)$, one axiom "for each truth degree $s$ and each truth degree constant $\bf s$ denoting it" since some truth degrees $s$ might not be denoted by any constant. I suspect that Gottwald meant here exactly what he wrote, so that, in the situation where $s$ is not denoted by any constant, there would be no axiom for $s$ in this axiom schema (just as, if there were three constants denoting the same $s$, then there would be three axioms for $s$ in this schema). It is, of course, possible that, in my quick glance at the earlier parts of the book, I overlooked some convention requiring every truth degree to be named by a constant, but I don't think Gottwald needs or presupposes such a convention in $Ax_{RT}6$.

Finally, you mentioned in your question a couple of specific $J$ formulas, built using negation and implication (and rendered hard to read by Polish notation). But at this point in the book, Gottwald is not assuming any particular construction of the $J$'s. He only assumes that there *exists* some appropriate $J_s$ for each truth value $s$ and that there is an implication connective (with appropriate semantical behavior). How the $J$'s are to be constructed will depend on what other connectives are available, and this might be quite different in various logical systems. The point of these axioms is that they permit proofs of some important facts about various logical systems *without* needing to go into the details of those systems --- a classic use of the (meta-)axiomatic method.