I have only vague advice for you. I have mixed GAFF and GLYCAM using a frcmod where I picked the best parameters I could find. However the accuracy required for the application was quite low (very simplistic modeling) so I could defend what I did. You'll have to figure that out based on what you plan to use the results for. You can describe the project here if you like. To do it "properly" you must develop charges as done for GLYCAM. I see you've been recommended here from this http://archive.ambermd.org/201811/0288.html
Xiao is probably best suited to answering on how to go about developing charges.