What are you comparing the results to?
Sounds like intuition regarding the design rather than hand calcs or physical testing. With that, I would suggest breaking your problem down into smaller chunks. Chunks that you have results for to gain confidence in the software ability to give accurate results. This will also help you understand how to setup the problem to get accurate results.
Without more info about the setup it is hard to guess at the problem. Here are potential issues:
1. Radiation. Room temp environment, 100deg components, might be important.
2. Materials. Are they real and correct?
3. Geometry. No problems...etc. check out blog.hawkridgesys.com on our recommendations
4. Setup. Are all the components and physics accounted for? Correct fluid parameters..etc. appropriate thermal resistances...etc
5. Convergence. Are you letting the solution converge?
The fact that you're seeing trends that you think are correct to me indicates you are on the right path but your setup needs work.