there are two ways that you can proceed:
1) if you want to count reads, then on the header line of a contig, use
say, to set the number the number of reads that map to contig055 to 666.
2) The latest version of MEGAN allows you to choose "read length" or "aligned bases" rather than "number of reads" as the number to be reported.
If your data contains both contigs and reads, then it would be best to use the "LongRead" algorithm for taxonomic and functional assignment, as this allows for the occurrence of multiple genes within any given contig.
However, we are currently doing a lot of work on the long read assignment algorithms, so that change from release to release, and so perhaps it is best not to use the long read mode until we have figured out the best way to do this.