I was indeed talking about the implementation and the algorithm as described in the article (prior to the recent edit updating the definition of L).

That Shamir's algorithm is correct was never in question.