I haven't read "The Selfish Gene" though I know of Dawkins' views. My original basis for understanding evolutionary thought comes from Gould, and you can see that influence here in my opposition to the view that behaviors have a direct genetic basis. Wikipedia summarizes his (and my) viewpoint nicely:
Gould's primary criticism held that human sociobiological explanations lacked evidential support, and argued that adaptive behaviors are frequently assumed to be genetic for no other reason than their supposed universality, or their adaptive nature. Gould emphasized that adaptive behaviors can be passed on through culture as well, and either hypothesis is equally plausible. Gould did not deny the relevance of biology to human nature, but reframed the debate as "biological potentiality vs. biological determinism". Gould stated that the human brain allows for a wide range of behaviors. Its flexibility "permits us to be aggressive or peaceful, dominant or submissive, spiteful or generous… Violence, sexism, and general nastiness are biological since they represent one subset of a possible range of behaviors. But peacefulness, equality, and kindness are just as biological—and we may see their influence increase if we can create social structures that permit them to flourish."
My resolution to the issue is to take the strict view that biological evolution only deals with genes, and that other mechanisms can be and are better at describing what you've termed the "higher levels" of selection.
The usual analogy is to computers, where software is constrained by the hardware, but still highly flexible. My own analogy is to think of weather. Fundamentally it's based on quantum mechanics, and at a higher level as particle dynamics, but weather forecast models use neither approach because because the minutia of those lower levels -- the detailed collision dynamics of a specific O2 against an N2 for example -- don't appreciably affect the large-scale weather. In fact, non-quantum mechanics can give rise to similar effects. There is no need to consider the effect of the weak force in order to predict tomorrow's weather.
Instead, weather is much more tractably understood as the gas law, along with parameters for "turbulent diffusion, radiation, moist processes (clouds and precipitation), heat exchange, soil, vegetation, surface water, the kinematic effects of terrain, and convection." It's absolutely true that weather is based in quantum mechanics, but it's not true that a quantum mechanics approach is the best way to understand the weather.
And I assert that that is the same for higher level behaviors, including even culture. Yes, there's an ultimate genetic basis somewhere in the depths, but a natural selection model based on genes is a poor instrument for understanding most human behaviors, and the attempts to do so have all had a sense of post hoc justification rather than having good experimental evidence.
Eg, just because certain dynamics models give rise to, say, homosexuality in a population, doesn't mean that those models reflect the actual process which gave rise to homosexuality in humans. It only means that gene-based evolution does not preclude an evolutionary model for homosexuality.
In any case, I gave a specific example of how an unfit behavior can spread to the entire population. It arises through a mutation, and that mutation happens to be in the same genome as another recent mutation which greatly improves fitness. This unfit behavior is both unfit for the individual and unfit for the group, but the negative consequences are so minor that there's little evolutionary pressure to weed it out.
I even gave the specific example of the endogenous retroviruses, which compose some 8% of our genome, which take some energy to maintain, and some of which are believe to have been inserted during a plague some 60 million years ago, though some are more recent. I find it hard to believe that any of those genes were ever beneficial to an individual or to the group, and yet after millions of years they still have not been fully weeded out.
(I do realize that these remaining sequences do not cause overt behavior. About all they do is take up time and energy during cell duplication. But that's still a behavior, and you have not placed a limit as to what evolution considers as a behavior or not.)