On top of that, we don't have radar tracking over the oceans. Transcievers don't communicate all the time while in flight. There are long periods of time where you are outside any national radar system. Compounded, it's entirely reasonable to lose a plane.
There are 15-20 planes that have never been found since the 1980s.
You'll also need to integrate that system across borders and jurisdiction which isn't easy politically.
It will also not surprise me if many actors want to have certain level of privacy as far as international airspace goes.
But the biggest factor at the end is cost compared to the relative costs of S&R missions if you did had additional tracking data. In the past 25 years only 3 scheduled passenger flights were not found including MH370 there were quite a few other private, chartered, and cargo flights that went missing mostly in tricky areas to search for in the first place like the Himalayas and the Indian Ocean.
It's doubtful that better radar coverage could even help with such cases even if you build a network of over the horizon radars which will scan for aircraft at various altitudes e.g. 2000-30000 ft. This will be very expensive and when you have a small aircraft that crashes onto a mountain side in an area where S&R can't get too or crashes over a 5KM deep ocean they might very well not help that much.
Probably the massive amount of money involved, and for very little gain. Most of them are going to crash hard and end up in the bottom of the ocean in little bits. Knowing where they crashed isn't going to save anybody, they're dead on impact.
But there's a lot to gain in analyzing the data. E.g. AF447 was a perfectly flyable plane that was lost due to bad piloting.
I don't know if Air France has actually started training and/or screening their pilots to help prevent a similar accident. But that's why we search so diligently; we need to know what went wrong so we can try to fix it.
Even if Radar will give you the exact point of impact with water (which it will never do) there are still so many variables that it might not be that much of a help to begin with over a general grid area.
An example would be a floating accelerometer/gyroscope/barometer combination found on almost all phones for instance. If plane crashes, the markers will float and its path can be reverse plotted based on stored sensor data, which we will get when we find one floating around. It doesn't need to have internet connectivity.
The ocean provides huge amounts of dilution.