Yes, lots of explosions and other events are filtered out. It's pretty easy to classify large explosions vs. earthquakes both by the depth (earthquakes are rarely so shallow; even a big earthquake that breaks the surface will have most of its energy released 5-15 km below the surface) and by the polarity of the first seismic waves (P-waves) that reach the instruments. In an explosion, all of the waves radiating out will have the 'first motion' being a push outward, away from the source towards the instrument.

However, tectonic earthquakes are rapid sliding movements of the crust along a fault that is a quasi-planar surface. This leads to patterns in the seismic waves, where half of them, two quadrants, show a compressional push away from the source and the other two quadrants have the first motion as tensile, or a pull towards the source. By looking at the geographic locations of the seismic instruments and seeing which show a pull and which show a push, seismologists fit planes separating the quadrants which correspond to the azimuth and inclination of the fault that caused the earthquake.

(edit: deleted ascii art which did not render correctly.

Interestingly, the orientation of the fault and the quadrants of push and pull reflect the directions of the stress field in the crust caused by plate tectonic motions, as well as secondary sources of stress such as gravitational loading of the crust from high topography, or the expansion and contraction of magma chambers under volcanoes. In the 60s through the 80s, geophysicists used this information to put together the first models of the rates and directions of different tectonic plates, but we now mostly use GPS for this.