You have two strict inequalities there: one comparing parameters, the other comparing variables. The parameters can be compared this way, but not the variables. So you could do something like:

set i / blah /;
alias (i,j);
parameter a(i);
variables x(i);
equation f(i,j);
f(i,j)$[a(i) > a(j)] .. x(i) =g= x(j);

Mathematically, we don’t do optimization over open sets (i.e. those defined with strict inequalities): how can you solve

max x
s.t. x < 3

And numerically, we do all this in finite-precision arithmetic so there is almost always some round-off error. Constraints are not satisfied exactly but to within some tolerance, so a strict vs. non-strict inequality is easily lost in the noise of round-off error.