Sensors in a standard S5 are also cooled by the fans, thus a lower reading, whereas in a water cooled setup, not much air movement is concentrated on the boards (aka sensors), so the readings are going to be higher.
In any case, if this were not the case, then we'd see that reflected in the other stats- hashrate, HW% etc, but we don't.
Cooling isn't there just for the chips, so if you're saying that its okay that the other components are too hot because the chips themselves are cool, you're gonna have a bad time.
To answer your other statement, we don't usually see temperatures reflected in those other stats anyway. Plenty of people using low RPM fans run at similar temps without any noticeable effect on the telemetry.
Another dogie-waffle response ...
The hashing chips are the primary heat generators in the S5, and the water cooling block (like the air cooling) is targeted at the chips. As was pointed out, the sensors are on the opposite side of the board and the reading they give in an air cooled setup is subject to the active air cooling, i.e the fans, which is not the case with a water cooling setup as the fans are mounted on the radiator (that should not be too hard to understand really).
If the temps are too high, you'll get HW errors when the chips fail to return a result (which will also reduce the reported hash-rate), and in the extreme case, the unit will stop hashing aka thermal cutoff .... and that is where temps are reflected in the other stats in case you were still trying to figure that out.