Complex Profile as Primary Datum (a "V" profile with a curve)

That sure is a tricky Primary datum to start learning with.  I feel for you.  Deep end indeed.   

Yeah, the software will only let you use simple 3D geometries as a primary datum.  Normally a plane, cylinder, or cone.   

A workaround for an unusual datum like that is to align to the feature with a Best Fit alignment and then construct an Alignment Plane that can be used for the datum definition.  Unfortunately, that isn't covered until the 201 course.  I'd like to explain how you can go about doing it on your part, but there is a lot of nuance that would be hard to quickly summarize for you.  I suggest you look at the Best Fit Alignment section in help file.  

Thanks for providing a picture.  That really helped clarify what you are dealing with.  Also, I'll just mention that the material modifier on the secondary datum of that other 0.040 profile is bogus.  Datum B is not a feature of size (like a diameter or a width), so that callout doesn't make sense

I think that worked!

To talk through what I did:

I made vector points across the entire Datum A profile, and created a constructed set of all those points. 

I did a profile tolerance on that set and it was great. 

I ran a best fit alignment: 3D, Vector Least Squares, and used all the vector points I used for the constructed set.

I then created a constructed profile and selected the 3 surfaces that make up the outside of the V (2 edges and the radius for the tip).

I then defined that as datum A

I then took 7 vector points on datum B, and constructed a best fit plane.

I did a profile tolerance to datum A and it looks good. 

I then took vector points along the last two edges and created and created a constructed set (again these are not perfectly flat). 

I did a profile on that to datum A and B and it also looked good. 

Thank you for the help! after I work on paths and making sure I have clearances I am going to try and run it (once I get the part)

Edit: For information purposes, this is how I did vector points the entire length of the part.

My thought process being I need two points on the flats for a normal plane, and 3 on the radius for basically a cylinder.  I did this formation 8 times down the length of the 40" part. 

I am hoping that defines the feature well enough. 

I'm wondering if a minmax vector bf alignment instead of LS could help to get a better datum - closer to a datum definition (just a question) ?

I'd be interested in what other people think of this too. 

Personally, I tend to prefer the reliability/repeatability of a Vector LS fit for a datum.  Though, I can see how some may argue that it is less correct to the design intent.  I certainly waffle on which is best to use as I encounter different types of part geometries.    

Really, I prefer that designers use datum target points on complicated contours.  It is just less confusing for everyone.