Ultimately, if the machine can use the saved password without specific human intervention, then... it can use it. This means that an attacker who gains full control of the machine (either an hostile software hijack with Administrator/root privileges, or he grabs the laptop and runs for it) will be able to recover the password. On a theoretical basis, if you want to resist such an attacker, then you need the password to never make it to the machine RAM itself; you need some tamper-resistant component which does all the processing. This is the model of smart cards and payment terminals, which have physical shielding and actively destroy sensitive data in case of physical breach; and they have no notion of "root user" accessible to the actual human user either.
At best you can hope to slow the attacker down. Yet, there still are management methods which are better than others. For instance, when you store some sensitive secret value, you want to avoid letting leak to the backups, or at least with some specific protection. Similarly, the user may want to integrate such storage within a bigger secret management solution, e.g. a "password safe" with a master password, or things like that. Your best bet, then, it to rely on OS-provided services for storage of sensitive data. On Windows systems, this points to DPAPI. On Mac OS X, use Keychain. As usual, the software which is easiest to design and implement and deploy is software which has already been designed, implemented and deployed by somebody else. Thus, don't try to do more or less smart things on your side; just let the OS do its work.