A lost language, spoken in Greece before the Greeks, surviving only as irregular words inside its replacement.
Documented
The ghost language is documented fact
Greek contains a large layer of words with non-Indo-European structure, identified by suffixes like -ss- (thalassa, Parnassos) and -nth- (hyakinthos, labyrinth, Korinthos), plus many place names across Greece and the Aegean. Beekes's etymological dictionary catalogs them systematically. Herodotus and Thucydides mention the Pelasgians, a pre-Greek population. Nobody knows what language it was, or what happened to its speakers.
Assessment: Fully documented as linguistics and as an unsolved identity. The substrate is as certain as the Indus Script's existence, and just as unreadable: a language known only by its fingerprints.
Stories
The Pelasgians who built the walls
The Greeks themselves told the story: the Pelasgians were here first, built the massive walls later generations called Cyclopean, because only giants could lift such stones, and then faded into the incoming tribes. The ruins kept the memory after the people were gone.
Assessment: Documented as Greek tradition and as archaeology: the Cyclopean walls are real Mycenaean and older work. Whether the builders were the substrate-speakers is unprovable, but the Greeks' own memory of predecessors matches the linguistic evidence of an earlier layer.
Interesting claims
The first Greeks were the newcomers
The strong reading: the sea words and mountain names being foreign means the Greeks arrived from elsewhere and absorbed a maritime civilization that named the Aegean first. On the fringe versions, that earlier people connects to Minoan Crete, to the Atlantis tradition, or to the lost-culture arguments on the Gobekli shelf.
Assessment: The arrival of Greek-speakers into an already inhabited Aegean is standard archaeology, not fringe. The identity of the substrate with any named lost civilization is speculation stacked on a genuine mystery, and the mystery is real enough to need no help.
Sources and further reading
Homeric non-Greek place names in -ssos and -nthos (Corinth, Tiryns)
E.J. Furnee, Die wichtigsten konsonantischen Erscheinungen des Vorgriechischen, 1972
Robert Beekes, Etymological Dictionary of Greek, 2010, the substrate chapter
Worth asking
An entire culture vanished so completely that only its loanwords remain, marked by strange -ss- and -nth- endings. What else has survived only as something harder to spell?