Acquaintance homicide is a broader category than one may think, and it shouldn't be excluded altogether from danger calculations. It is homicide in which the victim was known to the killer in any way prior to the crime.
The vast majority of "acquaintance" homicides aren't domestic homicides or homicides between friends. They are homicides among criminals, drug dealers and their customers or nearby residents, prostitutes and their pimps, and even a fair number of robbers and their victims. For example, per FBI reporting standards a criminal who enters a cab, rides it across town, and then murders the driver has committed an acquaintance homicide.
So one has a significantly greater chance of becoming a victim of an acquaintance homicide in an urban area even if he doesn't associate with criminals or have violent friends or relations. The drug dealer on the corner, a violent gang member known to you prior to an attack, etc., are strictly urban creatures.
Nonetheless, the study might still be correct about overall danger. The important point isn't the comparison to violent crime, however. It is the extent to which people die preventably on roadways. _That_ is a staggering amount, whether or not it's worse than crime.
It is also surely true that death by traffic accident is not random-- just as in cycling one can exercise a significant degree of control over one's risks on the road. The belief most people have, true or not, that _they_ are safer than average drivers likely would cause them to discount traffic dangers in their suburbs.