The water package is based on the steam tables. Please see here for most of the equations:
http://www.iapws.org/relguide/IF97-Rev.pdf(note that some revisions on backwards equation, particularly in region 3, have been used instead of the ones in this document)
Please have a look at eq 11, and particularly its deviation from the basic equation, which is given in eq 12. This is the deviation you are seeing.
It can be circumvented, I suppose, by refining the temperature answer that is obtained from the backwards equation by using an iterative approach, but this will surely make the code more complex (please note that this code also serves as example code) and the PH and PS flashes slower to solve (note that these implementations are based on the backward equations).
So let me ask you this, given that the deviation you see is the deviation that is present in the equations as presented by the steam tables themselves, do you experience this deviation as too large?
Please note that you can verify the implementation (
http://www.cocosimulator.org/down.php?dl=water_2.7.0.0_source.zip), as it is open source. Particularly, the IAPWS document provides check data for validating correct implementation of the equations, which you can find in the code (and all tests pass); this does not necessarily imply all answers are correct (there could be logic errors) but in this case I believe the answer to be correct, in the sense that it matches both the equations and the deviation as documented by IAPWS.
To answer your actual question: no - you did not configure anything wrongly.