Now that´s a statement.
As I wasn't really sure if he knows how to connect two meshes the usual way I somehow thought that it was necessary to make clear that I don't contradict you before mentioning how one could connect meshes. Sorry if I somehow managed to stand on your feet with that sentence, wasn't intentional in any way.
Aside from MeshMixer which I named already you´re probably talking about sculpting apps?
You probably mean Zbrush's Mesh Fusion, which is some boolean with fused together meshes in the end, no sculpting involved (with zbrush you can polymodel, but it's working different from the usual modeler apps). The result is quite clean, actually usable for hard and soft edge modeling and together with the auto retopo tools all-quad in the end.
Something similar is incorporated in Maya for years (and meanwhile in 3ds max, but I'm not 100 % sure). I don't know if it's better than Meshmixer or roughly the same.
Last time I looked Blender had at least a way to change the bevels after booleans (via scripts). Not quite the same.
Best in class (and the first one around) is still Modo's Mesh Fusion (Pixologic named their tool the same, but they didn't use the same plugin, instead creating something of their own). First this was available as a plugin (I think for lightwave, too), but it's for years integrated in Modo and further developed. It's something like live booleans where you can change the individual role of the involved meshes (add, subtract or intersect) and the seams (called strips). Actually, it's freaking awesome (but it feels like cheating). You can stick something together in minutes where you with traditional modeling methods need hours. The price you pay is a dense model because it works with subdivision. To a certain degree you can control the subdivision level of the different meshes, meanwhile from 0 on upwards, resulting in lighter meshes then before, but the result is still rather dense. With lots of meshes involved it gets a bit complicated, but you can create things from start to finish just with fusion and prefabricated meshes (some subd-optimized all-quad (not so) 'primitives').
The topology of the resulting Mesh is depending on what you feed it; often all-quad, sometimes some tris in the mix (again, you can use automatic retopology, with automatic guides). Usually you get a cleaner result with the traditional modeling methods. All in all it's better fitted for hard edge modeling, but with careful planing you can fuse together humanoid figures (including humans), very much depending on the parts you model. In the OP's case it wouldn't work out, though.
Sometimes, with such possibilities available, I'm asking myself 'why do I care?', especially as I believe that the days of the traditional 3d modeler (I mean the persons, not the apps) are counted. Stuff like this, procedural modeling, virtual reality and artificial intelligence together will result in ways how someone not knowing what a vertex is can create, texture and render complex 3d models (Adobe's Dimension is just a start), the actual work done by the computer (in the cloud, of course). My crystal ball has a crack, but in my view it's just around the corner (maybe 15, 20 years tops), as long as the climate, megalomaniac politicians with atomic bombs, religious fundamentalists with smaller bombs (of different faiths), total economic collapse and the general stupidity of our whole race don't stop the development of new technologies for a few centuries (or forever).