BLS doesn't look at job ads when compiling "job opening" data. Their method isn't perfect (nothing in life is), but far more comprehensive than you give it credit for.
The results of them actually talking to businesses and asking questions that are more than "did you have a job ad posted?" You are hardly the first person to imagine that job ads aren't representative of actual job opportunities. Obviously they are going to put in effort to avoid those weak signals.
Further, the graph shown is pretty noisy and I'm not sure the upward move which counters the downward trent is statistically significant.