High-resolution borehole strainmeters are usually installed in tectonically active regions in order to detect slow-slip events, and to estimate slow transients related to earthquake swarms. However, they are also sensitive to other numerous influences, internal or external. Furthermore, the quality of their coupling to the rock through cementation, and the mechanical properties of the rock mass around them, have a critical influence on their records. Many of the existing strainmeters present such problems, and the correction for these effects often remains a challenge. In this paper, we present the analysis of the records of a high-resolution borehole dilatometer (Sacks–Evertson), located in the seismically active rift of Corinth (Greece) (station TRZ in the Trizonia island). We show that the instrument suffers from an only partial-solid coupling, and that the nearby sea tides have a direct (through elastic response) and indirect (through pore-pressure diffusion) effect on the dilatation signal, which adds up to the solid tidal strain source. We propose a methodology that allows, in a first step, to better separate the internal (solid tide) from the external (air pressure, sea level) influences, by calculating a frequency-dependent transfer function outside the range of the tidal periods. We then extrapolate this function, in particular at the tidal periods. In a second step, the resulting variation with frequency of the coupling coefficients with sea level led us to estimate the proportion of instrument not solidly cemented to rock (thus in contact with water pore pressure), which is about 90 % of the total height. Despite the small proportion of solid coupling, the sensor resolution remains very good up to a few tens of hours of a time period, thanks to the confining effects of the rocks on the local pore pressure. These results allow us to correct for the external effects, and reduce the associated variance by 80–90 % (in the period range of minutes to days). The empirical correction of the sea level effect could be explained using a simple Boussinesq’s approximation and 1D pore-pressure diffusion model, which contributed to better constraint of some of the poro-elastic parameters in the vicinity of the instrument. After correction, the solid tidal signal at the 24-h period is almost anti-correlated with those of the theoretical solid tide. This surprising result is consistent with a similar anti-correlation observed for the longest period surface waves (200 s) comparing the TRZ dilatometer signals to the strain measured by a nearby borehole strainmeter (MOK, 15 km). This could be related to the presence of a shallow fault close to the instrument, which would creep in response to seismic wave-related stress.