My first suggestion was going to be drop the indexes on the target, so congrats on figuring that one out. Since you have two different sources you might find it faster to have 2 separate data flow tasks, one for each source. This may cause some locking contention on the target table, if you can partition the target table along differences present in the two sources you should be able to work out the contention issues. If not, I have also used a separate staging target (no indexes) for each source and then used an execute SQL task to assemble the two staging tables into the final target table. Even without 2 different sources you might find a way to fractionalize a single source by date or something to allow you to try this same approach.

A little bit of experimentation is in order here because there are a number of variables. But if you continue to experiment with indexes, partitions and/or staging tables I expect you can decrease your load times substantially. When I have experimented along these lines I have been able to drop load times from over 30 hours to under 5 hours.