TP1 and pin 2 of the header are directly connected, as are TP2 and the ground pins on the header.
So reading 4v across TP1/TP2 but 3.3v on the header means you're possibly miss-identifying the header pins.
Booting from only 4v instead of 5v is quite possible - on that early Pi, 3v3 was derived from a LDO linear regulator. If the input dropped that low, 3v3 would drop to out of regulation, but may work well enough for the Pi to boot.
Reading 2.5v somewhere on the header possibly means you're reading some unidentified GPIOs, unless the 3v3 rail really has dropped that low.
In which case, where is the 3.3v reading coming from?