Collisions in tag identification have a detrimental effect on the performance of RFID systems. Repeated queries prolong the identification time, consume bandwidth, waste energy, and can also aggravate collisions. A range of approaches aimed at resolving tag collisions has been introduced to increase recognition performance, and the combination of a knowledge-based query tree and bit tracking has yielded impressive identification results. In the combination approach, the knowledge tree plays a vital role because it exploits the static knowledge of potential tag identifiers stored in the database, which has a decisive impact on identification efficiency. Building on this foundation, we propose an improved Distinguished-Bit Tracking Knowledge-based Query Tree (iDKQT) that, at each collision node, forms an additional set of distinguishable tags from the remaining tags. This addition increases the number of separable tags per split, shortens the knowledge tree, and enhances identification efficiency while preserving the worst-case complexity order of the original protocol. In simulations with up to 1,000 potential tags under identical conditions, the improved method reduces, on average, relative to the DKQT: collision slots by 15%, idle slots by 45%, and total slots by 21%; and relative to the BKQT: collision by 37%, idle by 94%, and total by 40%. It also shortens identification time by approximately 18%, with a one-time tree construction that is approximately twice as long in static-knowledge settings. These gains are supported by a mathematical analysis that shows the added set increases the number of separable tags per collision without altering the worst-case complexity order.