The inability to navigate web pages linearly and intuitively using the keyboard, known as Linear Navigation Failures (LNFs), is a pressing issue for many users with disabilities who are reliant on Assistive Technologies that act through the keyboard interface. LNFs can lead to confusion for keyboard-based users when navigating through a web page and can hinder their ability to effectively access its intended functionality. LNFs are broadly unaddressed by current research, which is either incomplete regarding the various ways LNFs can occur or imprecise in detecting them. Therefore, we conducted an empirical study to identify the specific patterns by which LNFs can manifest in web pages. Guided by the observed patterns of these LNFs, we designed a UI modeling technique and a set of model-based heuristics to detect LNFs. The evaluation of our approach on real-world web pages demonstrated its effectiveness in detecting LNFs and that it significantly outperformed state-of-the-art techniques.