Manually Calibrated Bed Printing Tilted

After numerous frustrations with first layer issues, I took a long break from my Rostock Mini Pro and I am finally trying to get it working again. Effectively starting from scratch, I installed a fresh copy of Repetier v91, changed from 32 tooth pulleys to 20 tooth pulleys (80 steps/mm), switched to an E3D v6 hotend, and added an FSR leveling kit from Utlibots.

I have used the 120mm extrusion trick to calibrate my extrusion steps per mm (97 with an EZStruder from SeeMeCNC). I followed Minow's guide to adjust the screws on my endstops, getting my three towers all the same height. I then adjusted my delta radius and re-adjusted the tower end stops to get a flat surface.

Manually homing and lowering my hotend to Z5, I am able to move along the X and Y axis while remaining visibly parallel to the bed surface.

But when I attempt to print a test object, my bed seems severly tilted. Everything seems fine as I approach the X tower (left front), but the nozzle tears up the tape as I move towards the Y tower (front right) and then swoops up to 20-25mm as it hits the Y tower and turns back towards the Z tower (back center).

When I try to auto-level the bed, the X tower point is returning a value of around 5mm which I think is correct. But again the Y and Z towers are reporting more than 25mm of needed correction. Despite this, I still seem to be able to run my 4 calibration scripts manually to get a flat calibrated bed.

Just tried upgrading to v92 of Repetier and it is still happening.

Running a G29 to calibrate the bed returned an average error around 5mm which made me happy, but as soon as I started a test print the nozzle dug into the bed again and then started printing in mid-air. At first I thought maybe the nozzle drag threw the steppers off, but then the nozzle went back to the bed on the side of the print where the nozzle dug in.

Any help is appreciated. Thank you.

Eventually tried RichCattel's Marlin-Master fork, though most of the files are more than a year old so I am not sure I am using the correct branch. However I was able to successfully print out a 0.5mm thin wall calibration print as a test.

Unforunately the test following that print had massive extrusion skipping that I need to resolve now. It wasn't due to a clogged nozzle as I was able to push filament through manually once I removed the bowden tube, so I am going to try to increase the current to the stepper motor.

Also, sending a G30 A command with the firmware set to a 0.05mm calibration tolerance took approximately 37 iterations to calibrate, changing the diagonal rod length and radius which I thought I had manually calibrated better. Later switching back to the "default" of 0.03mm seemed to run indefinitely, so I finally killed the power to the board. Planning on switching back to 0.05mm (using FSR with PCB board).