The tables don't tell you what's better. They just give some estimates, which are the most likely values of coefficients in a model given the observed data. Note that these values tell us something we expect from (future, yet unmeasured) data and not about the parameters themselves (like RR etc). If the observed data is "problematic" for some reason, these estimates will be equally problematic. If the models don't make sense, the estimates won't make sense either. Only assuming that the underlying model does in fact make sense and that the observed data is in fact representative and free of severe "problems", one may take the given estimates as a first rough guess on reasonable values of the parameters.

The second problem is that Diosmin and Rutin are not compared directly. The two tables provide no clear information about their relative effects. Only when the differences between the two is really really large (practically), I would consider this some evidence for a differential effect.

And yes, a RR cannot be negative. It's unclear from the tables alone, but it could be that the log of the RR is given, what can be negative (a log RR of 0 would be neutral, positive or negative log RR indicate increased or decreased risk in the treated group, resp.