There are two issues that influence the room temperature measured:
1. the measuring principle: the UC… plastic cover perceives also the radiant part of the temperature, and changes it into increased temperature of the sensor body. If a black ball temperature sensor (such as http://products.domat-int.com/en/passive-temperature-sensors/178-pendulum-room-temperature-sensor.html) is used for reference purposes, a temperature close to what the UC… says would be measured. Most of the reference probes only measure the air temperature.
2. Another problem, of course, may be heat dissipation within the device. This is compensated by factory-set correction (which can be changed to fit the situation). But it already is set up so as to compensate the dissipation.
Please do check one thing: if the power supply voltage is higher than about 27 V, the internal overvoltage protection responds, and cuts its peaks. It may be a small voltage difference, but it is turned into heat, which then influences the whole inside of the UC… inclusive the sensor part. So an oscilloscope can be used to check if the power supply is a pure sine (OK), or a sine with cut peaks (not OK). An alternative is a thermal camera which shows if the heat is generated at the processor only, or also at the overvoltage protection element (a small component at the power supply teminals of the UC…).
Another reason may be the (warm) air coming from the piping in the wall where your cables are pulled, but this is not so probable. However, you can check if the pipes are sealed. We had such problem at one of the sites some time ago.