Once you consider that digital models doing the tone shaping it makes no sense to think of them as gain staged elements increasing output to a louder and louder level. All those models have an ideal signal level range to run in. Signals in the digital domain do not need to to be made louder as a whole, they are transformed in their frequency content and per-frequency volume by the models. Independent from an aggregate signal level all models have limits in respect to what they do when the output of their normal function exceeds what can represented by a number (e.g., when your Amp setting would duplicate the volume and you have a spike of input level at 60% the model simply cannot put out more than 100%, i.e., it must use some internal mechanisms to limit such a signal to less than 100%.
From what I measured there seem to be a -12dB magic limit in most of the Line 6 models (I have evidence for some EQs and all Amps). Once you go higher they start to soft clip (compress) and if you push them towards 0dBFS they distort more and more till they get into hard clipping mode(their internal processing can no longer limit the output to something less than 0dBFS and they simply cut off at 0dBFS. As far as I can tell the only exception is the mixer and possibly the volume pedal. They seem to do no soft clipping and simply cut when the output exceeds 0dBFS.
With that in mind one would probably want to take your guitar input signal as soon as possible to a healthy high level but not exceeding the -12dB at peaks unless you actually want some amount of clipping. All subsequent model stages should be focused on shaping the tone, but not to change the level of the signal unless changing the signal level is essential to achieving some intended effect (e.g., get some soft compression w/o an effect model as a side effect of going above -12dB output).
Good luck. Your millage may vary ;-),
Martin
PS: I gathered my data using USB and DAW for measurements.