Yes you are on the money. There are also some additional constraints, e.g., we don't want to match startups with multiple investors from the same fund.
In the NRMP there is a N:1 matching between residents and hospitals, where N <= the number of open positions in the hospital.
If the meetings are time sliced, let's say by hour, then run the algorithm for each time slice.
But regardless, I believe the algorithm is roughly the same as the NRMP algorithm: you run through multiple rounds where you look at each combination of company and investor. If both sides would prefer that they meet with each other, rather than the people they're already paired with, then you reassign them. Once you've done that several times you arrive at a stable state.