Frequency and ideal aerial length are closely related. At low frequencies such as 27MHz, a few centimetres either way on what is probably a piece of wire set to about 1/16th the resonant length doesn't make a huge difference, not really enough to notice. However, the odd millimetre out on something tuned to 100,000 times the frequency is going to be critical, and will probably prevent reception by effectively tuning the entire thing out of band. Real digital signals are tremendously robust - they can produce a good result even after a great deal of mangling, but when they go beyond limits, they GO. Very suddenly.
The big advantage of the lower frequencies in your situation is that you can fit a longer aerial, and you can mount it vertically. This way, there is less likelihood of it being hidden behind a wave enough to lose the signal. Bear in mind that signal strength does drop with range, and a partially hidden aerial will be less effective than one in full view. Also my previous comment about digital systems - none of the pre-2.4 systems are fully digital, they are analogue transmission systems that use the RX to create a digital signal. The RX can take a weak, possibly badly distorted, signal and produce a viable digital signal to work the servo, right up to the time when the limit is exceeded. Sensitivity will probably improve - selectivity will suffer (open the window further, more flies come in) but in your situation, this should not be a problem.
Most model boating happens at less than 400 yards range - we like to see the boat. On most venues the transmitter will be a good few feet above the water surface (and the bank usually doesn't go up and down by itself), so the very short aerials associated with 2.4, although very near the water level, are very rarely obscured. Many have two aerials, and take a vote on which is working, with a bias to doing something, which accounts for them continuing to work even in quite lumpy conditions. A long swell on open sea will be very different, not least because your boat will at some time be in a trough at the same time as the model.