I'm surprised that you are getting interference between the two sensors in this way. Have you tried sampling each sensor twice and discarding the first set of results. It may be that you are seeing an "old" set of results? So set both sensors into single shot mode, ping sensor1 <delay> ping sensor1 and use results <delay2> ping sensor2 <delay>ping sensor2 and use results <delay2>, repeat!
Would be interested to see what sort of results you get.
One thing I've noticed (and I think roger has seen the same), is that when running at high power you get a sort of geometric effect in the results. So if you get a reading at say 32cm you will also get a series of false results at some sort of multiple interval after that.
Finally you can also operate the sensor at 5V rather than 9V (versions of leJOS < 0.5 did this by mistake!). Doing this shortens the range, but also reduces the number of false echos you get. You can switch between 5V and 9V by using setType method on the sensor port TYPE_LOWSPEED is 5V TYPE_LOWSPEED_9V is 9V...