I just tested extracting reads to a new document and the feature generally seems to work fine. Please send me more details. Also, the latest release of MEGAN does support extracting reads from a comparison document.
However, please note that this is based on MEGAN accessing the original files that went into the comparison document. If those files no longer exist or have been moved, then MEGAN won't be able extract reads from the files.