You're not going to be able to measure (distance I assume?) on the blend because Pro/E doesn't know WHERE on that changing surface to measure, because you didn't tell it. In modeling mode you can query-select points on the surface. If you want dwg dims, put in a datum plane and use the interestion of that and the part to give you hard points to measure to.
An SPR is not needed, different/better technique is. 
For this, honestly, I'd use a VSS, and in the curves defining the shape you can put geometry points wherever you want to measure to, AND the VSS will give you better control anyways and assure tangencies where you want 'em.
Yup. And I do get the diameter, but would prefer to avoid dimensioning to hidden lines as that was pretty much beaten out of me in school. Yes, decades ago, but still remember the scoldings.....
.
And back to beating the dead horse: I should be able to pull a dimension where I want it, dagnabit! If the silly software(TM) knows that it's a diameter in one view, then it should in another!
A comparison for fun. The VSS is a little easier, I think, less sections to sketch, and you have more control over the geometry as well. What you can't do with a VSS is reverse curvature or switch between a straight line and curvature or vice versa.
Frank, there is a neat trick you can use to control tangency in the VSS without a guide curve.
I know this may require a math major to decipher, but in essence, wherever you can incorporate a cos() function, you can get a nice "soft start" on the VSS trajectory...
I made two cylinders and a "spine" that drives the shape. Everything else is covered by relations:
This is the VSS trajectory through the arc only; no blend option needed.
sd6=d3+((d6-d3)/2)-cos(trajpar*180)*((d6-d3)/2)
In simple terms, the sweep starts out 1.5" OD and follows a cosine curve which has a flat state at 1 and -1 (*180) and ends at 2" OD. (d6-d3)/2 manages the difference between the diameters d3 and d6.
In this case... small end =1.75-((1 to -1)*.25).
Now I have cylinders for dimensioning and a smooth cosine transition from one face tot he other without leaving it to the software, or a arbitrary chain curve (if that is not the control you wanted).
Nice! Thanks! I debated using a graph, but I figured that a guide curve was the simplest way to understand it for a novice user. I used a conic to cheat and make it simple, tangent, and reasonably smooth.
I really need to get better at math! 
Oh and don't forget that you have curvature control if you did this as a boundary blend between the 2 cylinders. Very nice if you need this to be cosmetic...
This is the cosine curve (same as a graph would be) for the trajpar statement above:
cos(trajpar*180)
This might help if you want to preserve the trajpar value of 0 to 1.
The cyan curve and related formula will create a "half height" cosine wave offset to invert it.
Play with the squared function by using cubed or higher power and you will get a variation between the start and stop.
The examples below are just the datum curve equations, but you would substitute the "y" entry for any "dn" dimension and replace "t" with "trajpar" in a VSS. Using the cyan equation in VSS relations will assure a tangent/normal start and finish but otherwise equivalent to the normal 0-1 range of trajpar.
Cool! I'll have to play with it. I'm a math moron so I avoid it and use geometry wherever possible!
