first
G1(x) = Emax(X - x, 0).
For an exponential with mean mu (so rate lambda = 1/mu) this is mu * exp(-x/mu). The expression previously here, exp(-mean * x) / mean, was written for a RATE parameterization while this class stores a MEAN: it put the mean in the exponent where the rate belongs and inverted the coefficient. The two agree only at a mean of 1.0, which is the default, so every default-constructed instance returned correct values and no other did.
Below zero the truncation never binds, because the support is non-negative, so G1(x) is simply the mean less x.