qProf sampling intervals varies wildly

[fgfletcher]

QGIS 3.34.3, qProf 0.5.1, Mac OSX 10.13/6 (High Sierra)
Project: EPSG:3614 NAD83(NSRS2007) / New Hampshire (ftUS)
Using a DEM with 5.5' x 5.5' pixels
Using a line shape file containing a single ~ 36,000' long RR track {i.e., gentle curves and slopes}
Setting "line densify distance" = 22 {i.e., 4 DEM pixels} to reduce # of points
I get reasonable statistics:

Statistics for profile # 1
Length: 35697.751784624925
Topographic elevations
- min: 426.71148553760935
- max: 720.9710075441237 elevations
elevations
min: 426.71148553760935
max: 720.9710075441237
mean: 581.3510160958499
variance: 7193.706459598757
standard deviation: 84.81572059234513
Sampling points (2324) for profile # 1 {averaging 15.36'}

The graph looks OK

However, when I export as .csv and look at the data, the spacing varies wildly.
I'm calculating the Euclidian distance between adjacent points, i.e., sqrt(∆X^2+∆Y^2).
Summing all those distances matches the reported length, so the math is OK
the spacing begins with
11.92, 22.00, 10.65, 22.00, 22.00, 4.73, 22.00, 22.00, 16.16, 22.00,
by midways I'm seeing
22.00, 2.25, 22.00, 0.90, 20.83, 18.03, 14.51, 18.82, 22.00, 7.75,
the last spaces are
11.28, 12.28, 13.03, 13.53, 13.79, 13.80, 13.58, 13.10, 12.40, 11.46

In another test, I used the same line, made new layers with the line shifted West 20' and another shifted East 20'.
This produced 3 layers, each containing lines of identical length and shape except for the sideways shift.
running qProf with the same settings on the shifted lines

the line shifted +20' results in 3014 points along 35,715.8', with point spacing averaging 11.85'
beginning: 11.92, 22.00, 10.65, 22.00, 22.00, 4.73, 22.00, 22.00, 16.16, 22.00,
ending: 0.18, 13.80, 0.26, 13.58, 0.36, 13.10, 0.50, 12.40, 0.69, 11.46

the line shifted -20' results in 2995 points along 35,679.6', with point spacing averaging 11.91'
beginning: 11.92, 22.00, 10.65, 22.00, 22.00, 4.73, 22.00, 22.00, 16.16, 22.00,
ending: 12.28, 0.04, 13.01, 13.46, 13.65, 13.58, 13.26, 12.67, 11.80, 11.11

These length differences match QGIS' "Identify results" values, but are rather unexpected given that the 3 lines are identical except for the shifting. I thought that, perhaps, the 3d distances were being used, but my 2D Euclidian distances exactly match, so that can't be the issue.

I was hoping to be able to classify the RR bed as being
above adjacent terrain
below adjacent terrrain
on a slope
by comparing the elevations 20' to the left and right of the centerline.

But the wild variability in the point spacing defeated my goal.

I am attaching the 3 layers so you can test them yourself.

Cordially,
Garth Fletcher
layers.zip