There are limits on what RAM can do, though there is room to be creative in combining RAM with shaping. For example, using shaping to funnel energy into one area that is a black hole to one set of wavelengths, and a clever combination of surface treatments addresses a finite set of others. This does not explain it well, but you may get the general idea.
As always it's the threat that determines your observables requirements. The B-2 was designed for a pretty wide range of threats as it had to persist in a very diverse threat environment. The F-117 and F-22 had very different requirements, as did TSSAM. So for a very diverse set of threats the B-2 does well overall, even if in specific bands it may not be as low as an F-22. The YF-118G Bird of Prey demonstrator did very, very well for threats comparable to the F-22.
Hypothetically.
If there were just one threat you were interested in - like HAVE BLUE against GUN DISH - today you could have a very, very, very low signature - but that may not be a realistic requirement. Just like anything else in engineering, it's a set of tradeoffs.
That is basically what LORRAINE was.
That's mostly true, but with some configurations there can be a higher degree of longevity - very large simple shapes like AARS and the original B-2, for example.
There are also opportunities to change the signature of the aircraft significantly without changing the outer mold line. Some current aircraft have been designed with this in mind.
