Thanks for the reply.
In a conversation with Tim Rajeff, he mentioned that (and forgive me if I mis-quote) line speed as he defines it is the velocity at which the upper half of the loop travels as a straight and taut "javelin-like" unit through the air. If this component of the cast is wiggly, flying upward, downward, sideways, etc - then no matter how fast the rod is moving it will lack speed over the course of the cast.
This way of looking at the loop has dramatically improved my thinking. I had worked my way to a very tight lower half of loop with a spey cast, and although both halves were tight with the single hander (probably more due to double hauling) there was much room for refinement. I have spent many hours tuning my stroke to produce the most efficient 'javelin' speed and shape I can. I have not perfected it yet but can produce the shape I want with increasing regularity each time out. Like yourself it has given me a basis for refinement and pursuit of a better cast.
Per the tailing loop, it sounds like you have already diagnosed the "shocking" of the stroke due to a late start. There is also the possibility that an upward drift is starting early on the other end, pulling the direction of the lower half of the loop upward while the upper half continues in it's original path - a collision course.
Bruce Richard's writings have also been of great influence to my practice regimen. He writes "Everything about the loop size and shape is determined by the path the rod tip takes while accelerating the line, and what it does during and after deceleration. Where the top leg of the loop is, and the top leg shape (curved up, straight, curved down, curved to the side) is determined by the path the tip of the rod takes while acclerating the line. The path the rod tip takes during the deceleration determines the shape of the "point" of the loop. What position the rod tip holds after the loop has been formed determines where the bottom leg of the loop will be.
If the tip path is upward through the acceleration, the top leg will be high, if the path through the deceleration is a rather broad curve forward, ending low, the loop "point" will be very round, if the tip ends up low after the deleceration that will direct the bottom leg low, resulting in a typical big round loop. "
Therefore in a tailing loop scenario, quite often the path of acceleration occurs in one vector, while the path of the rod tip during deceleration and position it holds after effects another. Just a thought, and interesting to think about. It might help you to try a slower, smoother acceleration going back - minding especially the path that the line accelerates on it's way to the start of the drift. Combined with starting the backcast just before your forward cast 'kicks' over you should be good to go!
Congrats on your progress!