Hi Antone,
Well I can't answer in a short way.
The Xmax is where the excursion is linear as a function of the input voltage. (not including thermal compression!)
This means, that while cone movement is Force x Springfactor and Force is B*LxI where B is the air gap magnetic field, I is the current in the Voice coil, L is the coil's lenght in the magnetic gap
This is why we use the B*L as force factor.
All this is true with constant B and constant Spring factor.
You can guess, neither of these is really constant in the reality.
However the mechanical parts (spider, suspension) are normaly designed that they are nearly constant in the region where the B is linear in the gap.
The B distribution can be modelled easily with some FEM softwares in these days, so we can guess it is optimized to be nearly constant in modern speakers. The only thing which changes now is the lenght of the VC inside the constant B magnetic gap, which varies with the amount of the excursion!
That means, Xmax one way(linear) is (should be in a good designed speaker):
(Hvc-Hgap)/2
Many manufacturers gives you the Xmax in the +0/-3dB B*L range, which means:one way Xmax:
(Hvc-Hgap)/2+Hgap*0,7
In this case the voice coil at his max excursion point is not 100% in the magnetic gap.
The Xmach comes from that where the mechanical parts limits the Xcursion for example bottoming the VC to the T-yoke or bottoming the cone/VC sealing to the top of the magnet assembly or the limit of the expansion of the spider/s or the surround.
The other big problematic of the speakers is the thermal coused resistance change, which completly changes the performance.
I hope this clears things
cheers,
Tamas