The original claim is "C++ is becoming a legacy language". This is clearly a claim about the trend of C++'s popularity over many years. You're trying to refute that claim with a single data point about one year, which isn't sufficient to refute a trend. For example, a single bad day in the stock market doesn't refute the claim "the stock market is trending up over the year". Moreover, your data point is probably just noise--a 4% swing when the error margins are closer to 50%. TIOBE is basically just guessing, and a guess isn't evidence.

FWIW, I've been a professional C++ programmer, and I like C++.