Looks promising...but Idle temps aren't really relevant though. The sensors aren't as accurate there as they are under load and so what you see at idle likely doesn't translate where it matters when loaded.
That is also, I'd guess, conductive... so some additional care should be used too...
I'd like to see this empirically tested under load instead of idle where God only knows what variables were inadvertent, different (mount, paste application, etc).