Civil liberties are most at risk in two circumstances: Bad faith and turf wars. One can argue either way about whether a single overarching intelligence directorate would increase or decrease the probability and extent of bad faith. On the one hand, it will probably be harder to avoid detection, just because so many others will now be working in parallel instead of in isolation. On the other hand, if/when it happens it will probably be closer to catastrophic than annoying. On the third hand, if there is a catastrophic failure, that's harder to write off as due to "bad apples," which therefore makes effective retrospective corrective measures that much more palatable and likely. On the fourth hand, that doesn't do anything for the particular victims. I've run out of hands; suffice it to say that these arguments are so speculative that they cannot be the key policy consideration.
One can't reasonably question the effect of turf wars, though. The "ordinary" justification for torture and less-obviously improper means of intelligence gathering is "we can't get the information any other way." All too often, that's just not true, if one looks at "we" as "the US." If, however, "we" means "this agency," it is. To use an unclassified example, consider the Vernona project (decrypting of Soviet intelligence and diplomatic communications during the early part of the Cold War). We don't know the CIA's (and DIA's) efforts to obtain precisely this information; nor do we know if there were any inconsistencies that tend to reveal a flaw in either the humint or sigint efforts. The one definitely should inform the other; the external indications are that they did not. If managers and control officers know that the information is available, though, they're less likely to think up and/or approve risky additional attempts to get that information through more-extreme measures, all portraits in spy novels to the contrary.
Compartmentalization goes only so far in the community; it has gotten extreme now. Consider, for example, one gaping hole in our counterintelligence/counterterrorism capability: Human intelligence on domestic cells. The CIA, NSA, and DIA aren't (at least in theory) authorized to operate domestically. Their mindset and training don't (or at least haven't been acknowledged to) consider that possibility. On the other hand, the agency that is authorized to act domesticallythe FBIis by nature reactive, not proactive, and has a pretty poor history of being able to penetrate close-knit operations even retrospectively. Thus, we'll have to develop an entirely new capability here; and pretending that CIA "observers" and "advisors" attached to FBI teams can fill the gap is at best wishful thinking, and more likely will only result in evasion of oversight responsibilities. Based on misuse during the 1950s and 1960s for partisan purposes, there are good reasons to keep the CIA, NSA, and DIA from doing much domestically; based on J. Edgar Tutu, one can even say the same for the FBI in this context.
Besides, maybe the spy novels will get better if we can start over with a new intelligence system and culture, so that the subplots don't all revolve around turf wars and rogue factions.