You can plot the TPXO grid (without land points with missing values) superimposed with your model grid to see how far your boundary is from the TPXO valid data. And I think you'll see that your model extend is too small.
You can look at the surface elevation results on boundary nodes where you prescribe tidal elevation. If the values are too low, it means that the model interpolates tidal signals from the TPXO database but located too far from your boundary. And it'll confirm that a solution is to extend your model domain.
In a general way, and location of your maritime boundary aside, the propagation of the tidal wave requires a relatively large domain to well represent it, so as the model can resolve the non-linear interactions between tidal constituents especially inside estuary where the tidal wave can change significantly. And it's your case I think. Don't forget to check your results with measurements if you can!
Once it'll be done, your model results should be better. You can play with CTIDE if necessary, but a constant value in time should be enough.
Regards,
Laurent