From previous experience, I found that I was making too many knowledge items while reading, which made it hard to identify the really important ones. So now, I read through the entire document (article PDF or chapter of a monograph) and mark down possible knowledge items (yellow highlight in citavi PDFs, pencil or 3M flags in printed materials). Then when I've finished, I write a summary knowledge item in 2 paragraphs—the first outlines the argument, the second (in italics) gives my evaluation and thoughts about possible uses. (I don't use the evaluation field in the contents tab as I want to be able to see this quickly in the knowledge view.) Now I'll go back through the marks I've made and work out which ones I want to input as knowledge items. As part of the input, I will assign to as specific category as possible. Every few days, I will have a look through my outline to see where the holes are (not a lot of knowledge items), and to see if anything in the NOT SURE YET category sparks any new thoughts on organization.