As far as I remember, baan will read the next n (20 or something) records in the buffer ...
So if you move to the next higher record it is already in the buffer and seems to be fast.
If you select an previous record, the querry will be executed once again ... which will take some time.
This was always a test for me checking the response time of the database and check if the statistics has to be updated.
But if you only have some data and running on oracle proper setting of your statistics should be checked ( we always set estimate statistics with 30 percentage).
To check if the oracle statistics are the reason you could switch the optimizer to rule based (not cost based) just for testing.
Another point could be setting proper values to max permutation (how many ways to the data should be checked before querrying ... we reduced this setting ... and get faster response).