This seems like the article's thesis to me. Here, the author argues that we should not seek to identify which authors/works are seemingly "worthy" of having an influence on other authors/works. Rather, we should explore on how literary influence is actually functioning in related works.Readers must look to different methods of influence, such as "distortion" and "misinterpretation" in their study of the topic. In the demands that she places upon her readers to be well-informed and attentive, Austen invites us to be a part of a complicated and ongoing literary conversation. Additionally, through studying Austen's works, we can observe the influence of those traditionally left out by the canon.
This argument does seem relevant and original to me. In my admittedly brief study of literary influence, the discussion is usually exclusively related to the canon. Murphy asks us to consider influence in a broader sense. However, the main question that I have after reading this article relates to computational literary study. Franco Moretti, Matthew Jockers, and other such scholars have made significant strides in the application of computational tools in the study of literary influence. I am very curious as to how this article's premises and main argument would hold up when subjected to such tools. This seems like a weakness to me. Even after my brief study of computational literary analysis, it seems that any conversation of literary influence is incomplete without actually looking at the data.