Curve around corners

[[To...]]

What is the best practice for measuring a curve that goes around sharp corners?

[[Ma...]]

I usually place one point in exact corner and make it 45° ( or any angle which suits there between faces ).

That will handle bad scanning ( mostly jumps ).
Inner corners i usually block evaluation for some points.
Also if you need that corner ( in your case ), then use smaller speed on some points ( make new segment with another strategy speed )

[[To...]]

While you're visiting my post, please remind me on how to do position of this irregular shaped pocket.  I always figured profile would handle that.  The above part has the following requirements below.

[[Ma...]]

That circle should mean "all around" - so not just one face, but all around that shape.

For TP - i would use either curve or that surfacce profile - use bestfit ( i would use only translation and rotation excluding in datum A axis )
Now place a point into nominal position as TP states and use element's alignment from that bestfit.

Now you can place TP on that point - it will move accordingly to bestfit.

[[To...]]

Please sign in to view this username.

 

That circle should mean "all around" - so not just one face, but all around that shape. That's why I called it "irregular shaped pocket"

For TP - i would use either curve or that surfacce profile - use bestfit ( i would use only translation and rotation excluding in datum A axis ) 
Now place a point into nominal position as TP states and use element's alignment from that bestfit.  The position is fully constrained. Doesn't that mean "no best fit".  2019 does not allow me to select a freeform surface or curve in my position

Now you can place TP on that point - it will move accordingly to bestfit.  see previous note

 

Edited 10 hours ago

[[Ma...]]

Profile is lock only for A - i assume perpendicular to A.
But TP is controled also by B and D and bestfit will be only for getting it's center. That center is then compared to ABD

Not for first profile callout.

[[To...]]

Please sign in to view this username.

But TP is controled also by B and D and bestfit will be only for getting it's center. That center is then compared to ABD  Sorry, I don't understand how to do what you're saying. Are you saying create a BestFit within the evaluation of the Curve feature with no translations or rotations? Then, how do I apply this to the point? Can you be more specific with the steps? 

Not for first profile callout. Yes, my question was only about position.

[[Ma...]]

To be safe you can make general surface and recall measured points into new one. This new one will have enabled bestfit with all translations and rotations -> save this bestfit.

Now this saved bestfit will serve as point element's alignment.

Let me explain it to you more deeply:

If you use a point alone, then you would have to somehow measure it, otherwise theoretical feature will report "0" as TP value.
So by using bestfit from surface profile will imitate real measurement, while nominal will be correct as from drawing.
So bestfit will move that point which is controlled by TP. By movement i mean like real measurement and you are comparing his real position to TED position in TP callout.