Walnut dieback can be caused by several fungal pathogenic species, which are associated with symptoms ranging from branch dieback to fruit necrosis and blight, challenging the one pathogen–one disease concept. Therefore, an accurate and extensive description of the walnut fungal pathobiome is crucial. To this end, DNA metabarcoding represents a powerful approach provided that bioinformatic pipelines are evaluated to avoid misinterpretation. In this context, this study aimed to determine (i) the performance of five primer pairs targeting the ITS region in amplifying genera of interest and estimating their relative abundance based on mock communities and (ii) the degree of taxonomic resolution using phylogenetic trees. Furthermore, our pipelines were also applied to DNA sequences from symptomatic walnut husks and twigs. Overall, our results showed that the ITS2 region was a better barcode than ITS1 and ITS, resulting in significantly higher sensitivity and/or similarity of composition values. The ITS3/ITS4_KYO1 primer set allowed to cover a wider range of fungal diversity, compared to the other primer sets also targeting the ITS2 region, namely, GTAA and GTAAm. Adding an extraction step to the ITS2 sequence influenced both positively and negatively the taxonomic resolution at the genus and species level, depending on the primer pair considered. Taken together, these results suggested that Kyo set without ITS2 extraction was the best pipeline to assess the broadest fungal diversity, with a more accurate taxonomic assignment, in walnut organs with dieback symptoms.