Not paste, although that's a possibility if improperly applied or the cooler isn't secure, AIOs work better at high pressure mounts. Most of cpu heat is in direct relation to the vcore voltage. A cpu running 1.3v is going to run considerably hotter than the exact same cpu under exact same conditions running 1.2v. Intel specifically sets voltages high for stock vcore simply to cover every cpu manufactured. So it wouldn't surprise me at all to see skylake stock vcore at 1.45v initial setting. This'll maintain a stable system even if the cpu is demanding 1.4v to run right.
That said, OC is more than just jacking the speed. OC is taking a stock cpu and changing how fast the cpu runs in relation to the stock voltage. So if you take a 3.4GHz cpu at 1.25v and change that to 3.4GHz at 1.208v you've technically overclocked since the cpu is running at a faster speed than the voltage was set for. This has the benefit of reducing temps. So when a cpu is bumped up in speed the temps rise accordingly, so lowering vcore will drop temps accordingly. Unfortunately the ratio isn't even, so while lowering vcore may drop temps 10°C, the additional speed has increased the wattage high enough that temps have risen 30°C, necessitating a larger ability cooler to absorb the wattage and lower temps. Therefore OC is all about max speeds for lowest vcore, just so the cooler can handle the difference.
Your temps are due to vcore. Drop the vcore until it's no longer stable (done in very small increments over many reboots and much stability testing) then bump it up a notch or 2. This'll set the base temp for your chosen speed. What comes after that is dependent on your hardware usage, software run, airflow, cooling ability etc but will no longer be dependent on the cpu or voltages as they are at minimum stable condition. Just because you have no wish to increase the stock speed doesn't mean you can't lower vcore. I'd start by returning the bios to factory default (usually F5) not doing it manually. Run your temp test, then start lowering vcore by 0.008v increments.
For temp tests use p95 v26.6 small fft, no newer version. For stability testing use Asus ROG RealBench. For Intel cpu's use Realtemp. It was written specifically for Intel cpu's for use with p95. With an AIO like the H55, temp testing will take a minimum of 1/2hr. It's needed to allow the liquid time to balance.