I have seen in many papers that Stream length gradient (SL) are classified based on relative tectonic activity into three classes, some are classified it as (class 1, SL>600; class 2, 300
The classifications of the statistical data sets are based on study objectives. Micro zonations can be addressed by narrow class width and vice versa. Although the subject matter is well addressed if the results displayed with spatial mapping applications like SURFER, ARCMAP, MAPINFO, etc. These mapping tools help in identifying the anomalous results.
In many cases it could be based on your understandings and knowledge about your unique case study area, and you can acquire this information from satellite imagery, field studies and investigation of active landforms in your study area. After that, in this method, you can use some software, for example ArcMap to classify your index in number of orders or categories as you want. It would be more compatible with reality. Furthermore, you can use some categories based on other authors and researchers studies (for example El Hamdouni, as you mentioned).