Table 1. Results comparison for different time limits applied to the Gurobi solver.

Four different time limits were evaluated to test the sensitivity of optimal solutions to the early termination criterion (60 s) imposed. In all cases, the λ-value was fixed to a reference optimum, the one obtained when the time limit was 60 s. Mean values for the 11 contexts (with the standard deviation within round brackets) are shown for the correlation between flux values and data, $\stackrel{-}{\mathbf{\rho }}$ _(V,D), the mean residual, $\stackrel{-}{\mathbf{R}}$ _(V,D), and the cardinality, i.e., number of reactions of the extracted models, $\stackrel{-}{\mathbf{C}\mathbf{a}\mathbf{r}\mathbf{d}.}$.

	30s	60s	90s	180s
$\overline{\rho }$(V,D)	0.4493(.068)	0.4493(.068)	0.4493(.068)	0.4497(.068)
$\overline{R}$(V,D)	0.2845(.081)	0.2845(.081)	0.2845(.081)	0.2844(.081)
$\overline{Card.}$	856.55(92.124)	856.27(91.153)	855.45(91.35)	856.55(93.17)