Sheeting a single alignment that has two pgl profiles

Glad you figured something out!  One of the keys to harnessing ORD as a team is file information federation, not only from a software efficiency perspective, but also sharing work among others (even yourself months later ;D) communicating design intent to others and maintaining as much "intuitiveness" as possible.  Breaking those out helps describe your intent, yet hold to standards that may or may not pass a compliancy report. 

There's nothing wrong with having two separate corridors for each direction of your design.  One corridor can point control to another corridor's template point (while each tied to the same CL stationing.)  In fact, you can setup your templates used by the two corridors to utilize two separate PGLs as vertical controls.  You do not need to use the profile a corridor is placed on as the template vertical control (it does require an active profile, but a floating template origin can make it possible to use a secondary vertical control.)  

You may verify you don't need two separate sets of BL stations if their profiles and super are independent (at least at necessary ranges where they will be different) .  You can still cut cross sections and associate design intent horizontally off the CL through that range, but your profile data should reflect those stations in addition to alignment BL displayed in planview for clarity. 

I agree with this.

We are required to show a number of profiles, and they all need to be annotated off the baseline. For the plan set we are required to produce, it doesn't matter that I am displaying the profile of the left EOP, it needs to be stationed along along the BL.

As an example, I consider a ditch line profile. We know that the ditch line will not be exactly as long as the baseline because, when the baseline curves back and forth, the offset ditch line will be longer or shorter than the baseline depending on whether it is on the inside or outside of the curve. The people reading my plans do not care. The ditch starts at station X+XX and ends at Y+YY, and the grade shown along the ditch needs to be based on that algebraic station to station length, not the actual length of geometry. I also need straight line segments, with defined PVIs - not a fuzzy projected profile. Let's make this even better, because I often need to show ditch profiles above or below datum for plans readability.

It sounds as if, to properly display this type of design data on my sheets, I need to define all of these profiles on the baseline. But doesn't that defeat the entire purpose of modeling?

Nope, because you can project those profiles from the baseline to each offset alignment you need to set a point control on.  Keep in mind when you project a profile with vertical curves the projected profile is a chorded line string version of your real profile on the baseline.  It's good enough for modeling though.