Agnostic biomarkers such as gene fusions allow to address cancer patients to targeted therapies; however, the low prevalence of these alterations across common malignancies poses challenges and needs a feasible and sensitive diagnostic process. RNA-based targeted next generation sequencing was performed on 125 samples of patients affected either by colorectal carcinoma, melanoma, or lung adenocarcinoma lacking genetic alterations in canonical driver genes, or by a colorectal carcinoma with microsatellite instability. Gene fusion rates were compared with in silico data from MSKCC datasets. For NTRK gene fusion detection we also employed a multitarget qRT-PCR and pan-TRK immunohistochemistry. Gene fusions were detected in 7/55 microsatellite instable colorectal carcinomas (12.73%), and in 4/70 of the “gene driver free” population (5.71%: 3/28 melanomas, 10.7%, and 1/12 lung adenocarcinomas, 8.3%). Fusion rates were significantly higher compared with the microsatellite stable and “gene driver positive” MSKCC cohorts. Pan-TRK immunohistochemistry showed 100% sensitivity, 91.7% specificity, and the occurrence of heterogeneous and/or subtle staining patterns. The enrichment of gene fusions in this “real-world” cohort highlights the feasibility of a workflow applicable in clinical practice. The heterogeneous expression in NTRK fusion positive tumours unveils challenging patterns to recognize and raises questions on the effective translation of the chimeric protein.