Sequences in GoldenGate uni-directional replication

Hi,

While testing uni directional replication with Goldengate, we are observing a strange behaviour with sequences.

For initial load, we did an exp/imp from source to target of the schema to be replicated. At this point we checked, and the last_number for a very frequently used sequence was same on both sides. We started the application and simulated some load and stopped after a few minutes.

Now the last_number on both sides is different; its higher by on the target side. In previous test the difference was 25. During every test, it is the target that is on the higher side. But why does this happen?