And the connection with Satellite's and GPS is ?

This is a nice misunderstanding. The fact that your mosfet can handle 2 W
does not mean that it does not heat up at 0,7W. Check the thermal resistance
from the datasheet, multiply that with your power (0,72W) and this gives you
the temperature rise above ambient. If that gives you 100C, the circuit is
ok and you just need to cool the mosfet. If the calculated rise is lower,
you either do not have sufficient drive voltage to the gate (mosfet not
completely on) or you have an unwanted oscillation at a high frequence which
can also produce a lot of dissipation.

Meindert

Did your calculations allow for the switching losses?

I've just been simulating a simple 1300 Hz PWM switch
where the simulation shows that the MOSFETs dissipate
about 2.5W each.  But only 1.5W of that is the "dc"
loss, another 1W comes from the switching losses. In
fact the simulation shows a peak of 160 V*I during
each on/off edge.

Sounds like inadequate heat sinking. This is an SO-8 dual MOSFET, and
it doesn't have real good thermal properties to begin with. See
Infineon's application note on thermal considerations with their SO-8
devices:
http://www.infineon.com/cmc_upload/documents/086/819/ThermalPackageConsideration
-DPAK-SO-8.pdf

If you are using the device without any provision for heat sinking,
you're going to see it run something close to 80K over ambient at 0.72
W. You need a low-thermal-resistance path to several square
centimeters of copper just to get the thermal resistance down to 60K/W
or so. If this isn't feasible, you need something in a different
package for which you can more readily arrange a proper heat sink.