I know you said you don't come around here much so I'll just let you know, this is better suited in the Random Questions pinned thread in this section
Anyways, I had similar results while over clocking the other night. It was idling at about 30*C then a little bit later it was showing 24*C.
I was using Real Temp to monitor the temperatures and before the temps dropped I ran the "Sensor Test" in the program. Then I opened up CPUID Hardware Monitor. Seeing the temps in CPUID HW Monitor showing mid 20's I went back to Real Temp and sure enough it was also showing these lower temps! I'm not sure if it's because I ran the Sensor Test or because CPUID HW Monitor did something... but both the programs were now agreeing with the same, lower temperatures with a higher clock speed and higher Vcore.
Those were idle temps, and running Super Pi to 8M for each 100Mhz increase I only hit 34*C (about 5 minutes of stress on the CPU to calculate 8M). And again, weird thing was the higher I went on clock speed, it started not going past 32*C in the same ~5 minute range of stress. I could see some reasons why it's not increasing temps, or at least not significantly but it still does seem counter intuitive.
I would like to warn you, it's generally not good to OC with such large increases. .4Ghz at once is pretty risky if you've never ran your CPU at those speeds and don't know it can handle it. If it ran stable, going from 3.4Ghz to 3.6Ghz isn't too bad, but I generally find that going .1Ghz is the most you should really do with out causing damage. But hey, to each their own.
So my recommendation would to take it easy on the increase in clock speed, and try out a different temperature monitoring program. I like CPUID Hardware Monitor, personally.
Also, make sure you're not over powering the processor compared to your RAM and mobo abilities, specifically the RAM.