In this paper we show how the performance of the joint hypocentre determina
tion (JHD) method can be improved, leading to reduced instability in cases
close to singularity. The method has been extended by imposing a number of
constraints introduced by other authors, and adding a new one. We tested th
e stability of the method and the relative advantages of the various constr
aints by simulating a geometrical space distribution of hypocentres recorde
d by a regional seismic network. We then applied this method to deep earthq
uakes that occurred in the Southern Tyrrhenian Sea subduction zone and to t
he seismicity of the Northern Apennines, which is moderately deeper than th
e typical shallow seismicity of the Apennines. The results obtained from th
e analysis of synthetic data and actual earthquakes confirm that the JHD me
thod produces less scatter in the hypocentral determinations with respect t
o the standard locations. The main conclusion of our study is that we can s
ignificantly reduce the systematic mislocations that result from applying J
HD to very clustered seismicity if we introduce the appropriate set of cons
traints.