Optical imaging and UV-visible detection of in-cylinder combustion phenomena were made in a single cylinder optically accessed high swirl multi-jets compression ignition engine operating with two different fuels and two EGR levels. A commercial diesel fuel and a lighter fuel blend of diesel (80%) and gasoline (20%), named G20, were tested for two injection pressures (70 and 140 MPa) and injection timings in the range 11 CAD BTDC to 5 CAD ATDC. The blend G20 has a lower cetane number, is more volatile and more resistant to the auto-ignition than diesel yielding an effect on the ignition delay and on the combustion performance. Instantaneous fuel injection rate, in-cylinder combustion pressure, NOx and smoke engine out emissions were measured. Taking into account the particular configuration of the engine, the efficiency was estimated by determining the area under the working engine cycle. Moreover, it is also known that the diffusion flame lift-off length strongly influences the soot formation, which is affected by in-cylinder gas density, air swirl and jet to jet interaction. To understand the role played by these factors on lift-off length and soot formation, images of UV-VIS chemiluminescence were used to measure the diffusion flame lift-off length. The results demonstrate as the G20 fuel blend, at late injection timing and high EGR (50%), increases the ignition delay, allowing to operate, at late injection timing, in a partially premixed low temperature combustion (PPLTC) regime in which the fuel is completely injected before the start of combustion. In this regime, strong reduction of engine out emissions of smoke and NOx were obtained with a penalty on engine efficiency. This limitation was overcome operating at earlier injection timing (11 CAD BTDC) in which a mixing controlled low temperature combustion (MCC) LTC regime was realized. In this regime, a good compromise between low engine out emissions and efficiency was made possible. Finally, images of UV-VIS chemiluminescence were also used to measure the flame lift-off length, defined as the distance between the injector orifice and the most upstream location of flame chemiluminescence on the fuel jet. Flame lift-off data were plotted versus the visible emission intensity at 532 nm, representative of soot emission. It was observed that, for all engine test conditions an increase of lift-off length corresponded to a decrease of in-cylinder soot production.