Interesting "Bug" In Sweep Curves (Wildfire 4)

Yesterday and today I set out to draft a cooling coil for a coworker.
It's a simple, flattened spiral of copper tubing lying on its side,
with two "risers" coming up from it (to come up and over the side of
the tank the coil will be sitting in).
Think oblate spring lying on its side, with its two ends pulled out
and running vertically upward, then bending over the lip of a tank.
The first riser descends to the bottom of the first coil, and then the
tube turns in 6 coils. The ends of the coils are basically 180-degree,
2-inch-radius bends of the tubing.
At the end of the last coil, the riser comes up off the top of the
coil, rises up the side of the tank, and over (shaped like a candy
cane).

I started out with a single set of Datum Points, using Offset
Coordinate System Datum Point and Cartesian coordinates, defined in
the order they would occur along the tube. The end of each coil would
consist of two 90-degree bends, and thus had two of these Datum
Points.
Each coil-end was progressively defined to be ~1 inch further in Y
than the previous end (those two points' Y-coordinate was 1" larger
than the previous two points) to get the spiral effect.
I then used this set of Datum Points to define a single Datum Curve,
using "Single Rad" and "Whole Array", starting from Point 0.
Then, I created a Swept Protrusion along this Datum Curve, using
"Insert", "Sweep", "Protrusion", "Select Traj" (here selected the
first segment of my Datum Curve), "Curve Chain", and "Select All".

Originally I drew my sketch (for the sweep) of the cross-section of
tubing so that it would start out on the inside of the curves of the
riser... however, this would put it on the outside of the curves of
the loops, making the tube itself have different bend radii depending
on if it was in a loop or in the risers, because the driving Datum
Curve was using only one radius, Single Rad.

So instead, I changed my sketch so that the sketch would sit 90 deg at
the side of the Datum Curve (90 degrees to the plane of all the
"bends"), and the plan was it would stay at the side of the curve all
the way through all the loops, thus keeping the same tubing radius
throughout the part. Here's where the bug appeared-- in practice, as
ProE drove the sketch along the length of the Datum Curve, after a
loop or two the profile started to "migrate around" the Datum Curve
until at the end, it had moved to the top/outside of the Datum Curve
instead of sticking to its side!

Of course, this resulted in the tubing's "bend radius" gradually
changing throughout the part, rather than staying the same as I wanted
it to.

If I make the Datum Curve use multiple radii instead of Single Rad,
and put the sketch back on the inside or outside of the turns, this
"sort of" fixes the problem-- but Pro/E still restricts my choice of
radius to the max for the *smallest* turns... so the bigger turns look
more like rounded corners than actual 180-deg bends of tubing.

Any suggestions?