score:4
There were a couple of islands in Laptev sea. Semenovsky and Vasilievsky. They are present, for example, in wikimapia (the mark is in russian, it says melted Semenovsky island).
Wikipedia states:
Storms and currents due to the ice thawing significantly erode the islands, so the Semenovsky and Vasilievsky islands (74°12"N, 133°E) which were discovered in 1815 have already disappeared.
So answering your question
when did phantom islands stop disappearing on maps
they never did and never will.
Upvote:2
A modern example is New Moore/South Talpatti island on the border of India and Bangladesh. Google Maps shows it but Bing Maps does not.
It seems to have been a sandbar which appeared in the 1970s but was washed away at a later date. The surprising thing is that although it has been subject to a major boundary dispute and although the surrounding Sundarbans mangrove forest has a population of several million, its appearance was first reported several years later by satellite and its disappearance first reported by a climate change researcher an unknown time after it stopped existing.
Upvote:3
Comets are probably the most-modern equivalent to Phantom Islands of old - given that astronomical instruments are 'only so good' at keeping track of them (and similarly-orbiting space rocks)
There is also a pseudo-'phenomenon' of online map errors, which could be seen similarly to phantom maps of old. Think of this especially in light of how map providers remove real objects for national security.
Upvote:3
This may not count as "modern", but its a good story along the vein you are looking for ...
In July of 1610 Galileo was still making discoveries faster than he could publish descriptions of them. On the 25th he discovered that Saturn was apparently situated between two smaller companions that always moved together. Wanting to establish his priority of discovery, but not yet ready to reveal what he had found, he sent to Kepler (and others) the following jumble of letters, which he informed them was a coded description of his latest discovery:
smaismrmilmepoetaleumibunenugttauiras
Encoding dicoveries that weren't quite ready for prime-time like this was apparently a common technique back then.
Returning to Kepler and his struggle with the mysterious coded message, it so happens that after a great deal of effort he actually succeeded in making sense out of Galileo's jumbled string of character:
Salve umbistineum geminatum Martia proles.
which means "Be greeted, double knob, children of Mars." In other words, Kepler deduced that Galileo had in fact discovered two moons of Mars...
Kepler had already deduced Mars might have two moons, as Earth had one and Jupiter had four, so two would fit in the geometic progression. (Hey, it was the 1600's. Give the guy a break.) I suspect this is what he wanted it to say, so he found a way to make the anagram say that.
For years everyone just went around incorrectly thinking these scientists had established that Mars had two moons. People like Voltaire and Swift even wrote about them. Then about 250 years later, somebody with a far more powerful telescope looked at Mars and discovered that it had...wait for it...two moons!