In my mind, any database representation should reflect only the data in the original documents. A database should collate data. Assumptions can be applied to data but not be loaded into it. Otherwise it can give a false idea of "truth" which can only be what is in the original documents.
I agree Andy & Guy's posts are good ways of firming up your assumptions for summarising data. Many parish records include places of residence this would also validate supposed connections.
Thank you for your comments, and to Sillgen for hers.

I, too, now have the view that I cannot make any assumptions (however logical and valid they seem to be at the time) as if I cannot absolutely guarantee that the Adam & Eve who had a child in 1831 are the same Adam & Eve who had a child in 1833 then I risk missing valuable links or skewing the entire database. So, the (possible) duplicates go in then!!!!

Incidentally, Boongie Pam, I would be interested in talking to you about how you are dealing with a similar type of large scale database....
Thanks again to all who have replied and set me on track for many hours of filtering out the duplicates
