Hm, with cool Jmp features like a hierarchical X axis, stacked plot  and Label by % of Factor ... 
to recalculate the label values and positions ...
I think I will live with the unnecessary values 
Or wait, cheating a bit, using @Jed_Campbell's reply to the third question 

 
Open( "$SAMPLE_DATA/Titanic Passengers.jmp" );
Graph Builder(
	Variables(
		X( :Passenger Class ),X( :Sex, Position( 1 ) ),
		Overlay( :Survived )
	),
	Elements(
		Bar(
			X( 1 ),	X( 2 ),
			Legend( 12 ),
			Bar Style( "Stacked" ),
			Summary Statistic( "% of Factor" ),
			Label( "Label by Value" )
		)
	),
	SendToReport(
		Dispatch({},"400",ScaleBox,{Legend Model(12,
				Properties( 0, {Fill Color( 36 )}, Item ID( "Yes", 1 ) ),
				Properties( 1, {Fill Color( 19 )}, Item ID( "No", 1 ) )
			)}
		)
	)
);
(Current Report() << xpath("//FrameBox")) << {Text Color( "Medium Dark Red" )}