Hi Bonnie,
You could always use a peak of 0.5 so that when the peaks of two GAM functions are in close proximity they sum to slightly greater than 1 instead of slightly greater than 2. Provided you use the same peak for ALL subjects this shouldn't make a difference when it comes performing group analyses.
If you calculate percent change scores having a max of ~1 makes interpreting them easier, even though they're slightly wrong (the peak is > 1). But since it's slightly wrong by the same amount for all subjects it shouldn't be a problem when it comes to performing group analyses.
Anyway, this works for me, but your milage may vary.
Bye,
Colm.