One problem working with long reads is that erroneous insertions or deletions break the frame of translated alignments. The solution is to use a frame-shift aware aligner such as Last or DIAMOND (which is more suitable for working with MEGAN). However, if you want to use long reads or assembled reads with tools that use translated BLAST, such as CheckM or Prokka, then you need to first fix the frame-shift problems. MEGAN provides new methods for exporting your long reads and contigs as frame-shift-corrected sequences. Either use the export menu item File->Export->Export Frame-Shift Corrected Reads or use the new command line tool tools/read-extractor (as of 6.14.0).
For more details on analyzing long reads and contigs, please see our pre-print here: BioRxiv