Abstract Transmission trees can be established through detailed contact histories, statistical inference, phylogenetic inference, or a combination of methods. Each method has its limitations, and the extent to which they succeed in revealing a ‘true’ transmission history remains unclear. Moreover, the net value of pathogen sequencing in transmission tree reconstruction is yet to be assessed. We explored the accuracy and sensitivity to biases of a range of methods for transmission chain inference. We studied eight transmission chains determined by contact tracing, each one having more than a third of its cases sequenced (87 samples over 199 cases in total). We compared three inference methods on the selected transmission chains: (i) phylogenetic inference: the Ebola virus (EBOV) sequences derived from patients were mapped onto a dated EBOV phylogeny tree including 398 EBOV sequences sampled in Guinea between March 2014 and October 2015; (ii) statistical inference: we used the maximum likelihood framework developed by Wallinga and Teunis to infer the most likely transmitter-recipient relationships from the onset dates; (iii) combined method: we inferred probabilistic transmission events using both pathogen sequences and collection dates with the R package Outbreaker2. The cases coming from each transmission chain were mostly clustered together in the phylogenetic tree. The few misclassified cases were most likely allocated to the wrong chains of transmission because of the timing of their symptom onsets. Probabilistic transmission tree using only onset dates broadly matched the contact tracing data, but multiple potential infectors were identified for each case. The combined method showed that an a priori knowledge of the number of independent imports had an important impact on the outcome. Although cases were allocated to the correct transmission chains, discrepancies were found in identifying direct case linkage and transmission generations within a chain. Phylogenetic, epidemiological, and combined approaches for transmission chain reconstructions globally concurred in their output. Sequence data proved useful (if not necessary) to place the sampled cases in a wider context, identify transmission clusters, and misclassified cases when epidemiological chains are inferred from date of symptom onset only, and to identify links between supposedly independent chains of transmission.