Presumably the theory of subduction (plates sliding beneath other plates and causing mountains to form) is still a theory – and has not been verified (or observed) which is why it has now received the modelling treatment in earnest. A paper in Nature Geoscience (Nov 2016) seeks to show carbon in the Mantle is derived from subducted crust – see http://phys.org/print397921452.html
There are currently several theories on how carbon came to be in the Mantle – and the most obvious one is that it is of primordial origin, formed during the creation of our planet. Another idea is that carbon in the Mantle came about as a result of a collision – such as the one that disgorged the material that became our Moon. Another idea is that it is derived from the marine environment. There is a lot of dissolved carbon in sea water and vents on the bottom of the sea floor, or submerged volcanoes come to that, may play some kind of role. In the new paper it is thought carbon may subduct in the crustal material that is swallowed up at plate boundaries (where one plate overrides another causing one to fold and the other to sink into the Mantle. You can take your pick. However, the latest exercise is done by modelling – on the assumption that subduction is a process that is theoretically correct and as a result of the import of data based on that belief the conclusion is signed and sealed. What does it really prove – that subduction takes place or that a computer simulation is biased?