The matching network can't improve the antenna bandwidth.
Do the antenna design correct, and no matching is needed was previously my opinion. Was mainly working with fixed external antennas.
For an embedded cellphone antenna, is that not an relevant opinion, but these did not exist then.
Started develop embedded antennas in late 90s.
Correct used is impedance matching a part of the embedded antenna solution. Even if antenna impedance alone is rather extreme can antenna and matching network together result in relative good antenna performance.
In a smartphone, there is not space enough, there is not height enough to make a conventional antenna structure with VSWR below 1:10 in 5 or more world-wide 2/3/4G bands and with efficiency above 40% in all bands.
It was many years ago it was possible to implement a dual band PIFA in a cellphone.
When I design a multiband cellphone antenna, I must admit that I sometimes not even know antenna impedance until afterwards, when complete solution antenna+matching network design is finished.
Focus at development phase is for best efficiency for actual frequency range, do not care anything at all about antenna impedance as I use a software that constantly add a optimized virtual network during design process.
My instruments shows resulting efficiency and impedance for antenna+network as that is what I am going to implement in the smartphone.
Not really any complex antenna design, but at
https://www.antune.net/demo/bluetooth/index.html do I match an BT/WiFi antenna where antenna structure alone have a VSWR 1:50 as best. That is no bandwidth at all. Antenna size is 1*7 mm and with 1 mm isolation from ground.
Final antenna+matching network had a measured efficiency around 45% for a bandwidth covering 2400-2500 MHz with VSWR below 1:3.