Myocardial fibrosis is a key structural feature of dilated cardiomyopathy (DCM) and alters ventricular activation, however reduced-order models used for ECG fitting usually model it into a globally reduced conduction velocity. We investigated whether representing fibrosis explicitly improves the approximation of the QRS complex in two patients with idiopathic DCM. Patient-specific fibrotic regions were segmented by a clinician from late gadolinium-enhanced magnetic resonance imaging (LGE-MRI) and included to a reduced-order Eikonal model as a distinct, slowly conducting region whose velocity was inferred alongside the tissue velocities and earliest-activation sites. We compared four personalisation scenarios: a reference solution without fibrosis, explicit fibrosis modeling, and two variants in which the reference conduction velocity priors were globally scaled down to 80\% and 60\%. Explicit fibrosis improved the fit over the reference in both patients, raising the mean Pearson correlation coefficient (PCC) from 0.920 to 0.947 and from 0.885 to 0.895, respectively, and reducing the root mean square error (RMSE) in both. Global scaling down the CVs reached a higher PCC, but depended on the chosen factor: scaling to 60\% degraded the fit in one patient below the reference. The inferred fibrotic velocity (0.11 and 0.18~m/s) was weakly constrained by the QRS.These findings suggest that explicitly modeling fibrosis provides a more physiologically consistent representation in DCM and improves ECG based inference, supporting its inclusion in reduced-order cardiac digital twin models.