Typically, the first run of the miner takes longer than usual. This is likely because it is parsing through queries beginning from the date chosen during setup.
We recommend that the start date be no further back than a week. As long as the miner is scheduled to run and running, it will continuously pick up and build lineage as data flows run.
Subsequent runs should be much quicker in relation to the first run — especially if the miner is set up to run daily. In that case, it would only be parsing through new queries as opposed to historic ones. Keep in mind that the number of queries or transformations running daily can also be a factor in the time it takes for the miner to run.