Two types of complex nanotubes produced by the arc-discharge method are investigated in this study: multi-walled carbon nanotubes filled with metallic nanowires and composite BN–C nanotubes. Their multi-element character yields specific spatial chemical arrangements — deduced from transmission electron microscopy (TEM) studies — which give precious information about the growth mechanism of nanotubes. Concerning filled nanotubes, if the metal–graphite cathode is carbon free, no filling is obtained, whereas if it contains sulfur — either as an impurity of graphite or when added under controlled quantities — complete or very long fillings are achieved, even for metals with very high melting temperatures. The chemical analyses revealed various types of fillings: pure sulfides, grains of sulfides alternating with grains of pure metals, or in some specific cases, pure metals. As for the B–C–N tubes, a total phase separation is observed between BN and C, and these two phases form concentric shells typically of the C/BN/C type. In this paper, we show that an approach combining the vapor–liquid–solid (VLS) scheme and the characteristics of the solidification given by phase diagrams account very well for the observed structures. The contrast between the absence of filling, when a sulfur-free carbon–metal rod is used for the cathode, and the successful fillings, when sulfur is added, is explained by metal–sulfur phase diagrams: adding sulfur to a liquid metal decreases the solidification temperature. The different types of fillings are also explained by the nature of the sulfur–metal phase diagram. An eutectic solidification, such as in Ni–S, yields two phases — the pure metal and the first sulfide — within a given tube, whereas the existence of a miscibility gap in the liquid, such as in Cr–S, leads to two separate liquids and, therefore, to two different fillings. In the same way, the eutectic-like pseudo-binary C–BN phase diagram explains not only the complete phase separation between BN and C, but also the observed organisation between layers: we propose that the latter is due to a sequential solidification of the two phases. As a perspective, this phase diagram approach is also discussed in the context of the formation of ropes of single-walled carbon nanotubes from the solidification of a metal–carbon liquid particle.