Ok, this is an old thread, I've just been reading through them and wanted to add to this one if anyone else comes across it some day.
The answer is yes, on average. Normally though we aren't dealing with average when human selective breeding is involved, as normally the exceptional is what is selected.
Diploid liliums have 24 chromosomes (12 from each parent) and each pair will have a large number of genes fight for dominance.
Say the majority of trumpet height gene/s is on chromosome 3 (simplifying, height is likely governed by multiple genes on multiple chromsomes, this is just a simple example for traits that are immediately obvious and being actively selected for) and it's dominant (or at least co-dominant given that it's expressed in the original O x T cross). You could theoretically breed it 5 generations (so 98.4375% oriential and 1.5625% trumpet) and still have that chromosome 3 (height) being from the trumpet (which realistically means even if only that 1 chromosome 3 from the trumpet 5 generations back is all that remains of the trumpet that's still 4.166667% trumpet rather than the 1.5625% that averages would suggest). Technically you could breed it infinitely if you're actively selecting for it and it's dominant. The 50/50 split, averaging, works in the wild or when considering each generation seperately, not so much when selectively breeding over generations since that selection actively skews results.
Sure each generation makes the desired combination of chromosomes passing on less common (to a point, going back to height dominant on chromsome 3, if that's all you're selecting for then at worst if you only breed it to orientals, on average you're going to be looking at 25% of selectable offspring each generation, if you're actively searching for dominant traits from 4 different chromsomes then it goes down to an average of 3.125% being selectable which isn't that bad but highlights you probably want to start breeding more within a slightly more related pool if those traits are important) but that comes down to human selection (and possibly repeated breedings until the dice roll the right way).
So on average yes, but human selection can dramatically skew it based on selection criteria (and possibly persistence and luck).
