What is the main point of the arguments in "On Actor-Network Theory: A Few Clarifications" by Bruno Latour?

Latour's article seeks to correct some misconceptions about Actor-Network Theory (AT) as described by Callon, et. al. First he concerns himself with the problematic nature of the word "network." Latour makes two points:
1. AT networks do not necessarily bear any resemblance to technical networks such as computer systems. AT networks can assume many final states, and the rigidity of computer networks is only one of many possible outcomes.
2. Nodes in an AT network have little to do with social networks, and in fact, the nodes of AT networks are very likely to be non-human or non-individual actors. In this way AT is less interested in describing a discrete network than in "the very essence of societies and natures."
AT is a "change of topology" in which networks are understood as multidimensional arrays of "filaments" or "rhizomes." The strength of an AT network lies not in "concentration, purity and unity" but through the "netting, lacing, weaving, twisting" of filaments into a robust whole.
Latour continues by describing the "simplest properties" belonging to all networks: that geographic distance is irrelevant or of less importance than the connections between nodes and that physical scale is irrelevant as well ("A network is never bigger than another one"); also irrelevant are notions of "inside" and "outside ("A network is all boundary without inside and outside").
Latour next describes the three "preoccupations" of AT, which are 1) "a semiotic definition of entity building," or the semiotic investigation of the narrative nature of networks, 2) a methodological framework within which 1) can occur, and 3) a statement on the ontological nature of actants. Just as semiotics is a way of describing narration, AT is a way of "recording" the "deployment of associations." By "taking the meaning bit" out of semiotics and understanding it as a way of describing the order of things (or "path-building")—not simply texts, but objects as well—texts and meaning are raised to the "ontological level of things."
In this way, AT is not "about traced networks but a network tracing" activity. AT does not seek to anthropomorphize networks or actants (a common misconception is that actors in AT always function like humans). Instead, it is about "what moves" in networks and how this movement is recorded.