I assume megablast is different that running a discontinuous blast (dc-megablast)
This is taken after running blastn -help
-task <String, Permissible values: 'blastn' 'blastn-short' 'dc-megablast'
'megablast' 'rmblastn' >
Another source of variation could be the database used. The local one cannot necessarily coincide with the remote if the newest sequences are not included. I believe they update the databases at a monthly pace