Start with your endpoints, assume range of corey exponents (oil wet & water wet) - put into your model and find the ones that come closest to matching your history

The problem is that you state you can't trust your SCAL, so you really are reduced to retrofitting (history matching) a set of curves that yields appropriate behaviour. Note: Depending upon aquifer strength, proximity & volume of production, just because you've produced mainly oil does not necessarily mean you have not got an oil wet system - it could simply mean that water is too far away, or you haven't produced enough volume to bring it in yet