No, the former. The quadratic extension numbers are always normalized, having square roots in the numerator.I am not fully certain whether this expression is -1/2 + (1/2)r5 or -1/2 + 1/(2r5) , although I suspect the latter.
You are cordially invited to extend the interfaceTOSimplex can return all feasible solutions for the integer variables of a MILP, but this feature is missing in the interface to polymake