Double-black-hole binaries are among the most important sources of gravitational radiation for ground-based detectors such as LIGO or VIRGO. Even if formed with lower efficiency than double-neutron-star binaries, they can dominate the predicted detection rates, since black holes are more massive than neutron stars and therefore could be detected at greater distances. Here we discuss a new binary evolution process that can very significantly limit the formation of close double-black-hole binaries: the vast majority of their potential progenitors undergo a common-envelope (CE) phase while the donor, one of the massive binary components, is evolving through the Hertzsprung gap. Our latest theoretical understanding of the CE process suggests that this will always leads to a merger, preventing the double-black-hole formation. Using population synthesis calculations, we find that the corresponding reduction in the merger rate of double black holes formed in galactic fields is so great (by \~500) that their contribution to inspiral detection rates for ground-based detectors could become relatively small (~1 in 20) compared to double-neutron-star binaries. Our predicted detection rates for Advanced LIGO are now much lower for double black holes (~2 per yr), but are still high for double neutron stars (~40 per yr). If double black holes were found to be dominant in the detected inspiral signals, this could indicate that they mainly originate from dense star clusters (not included here) or that our theoretical modeling of the CE phase needs significant revision.