Rigid tap on Lathe small bounce in Z after retract

I tried rigid tapping on my lathe for the 1st time and was hoping to get a bit of advice on how to stop the bounce I'm getting in the Z after the tap pulls out. I have a start and return position of .5 inches. It looks like synchronization is still active past that point as the spindle reverses and as a result pulls the tap forward again for a split second. Do I need to add a dwell somewhere? Is exact stop mode preferred or path blending with tolerance have any influence on tapping?
RPM was 500 at .050 lead.1/4-20 tap. Mesa hardware, full encoder on spindle and running axis on 2.8.0 pre1-3669.

BTW, I did break 2 taps in the process (a nice new spiral flute) . Not related to the retract bounce. I think the tapping depth for a blind hole was a bit much at 1.060" and I didn't oversize the hole enough (.201). I'm using a compression/tension collet with a small amount of float just in case. Material is 1215 steel. If anyone has experience with deep blind hole tapping please chime in.

Nice video

According to the doc that is the way it works. Synchronized move during retract until spindle stop. Then fast forward to start position - this is the bounce you see. To reduce the bounce means reducing the time for the spindle to change direction.

A floating, flexible chuck could save the tap. I have no experience with that myself, but have been looking to acquire one.

Thanks Haken. It looks to me the tap pulls forward for the split second the spindle reverses from M4 to M3. Why does it remain sinchronized past the point of start/end position?
I find the explanation in the docs confusing. From the G33.1 section my comments in red:

1.A move to the specified coordinate, synchronized with the spindle at the given ratio and starting with a spindle index pulse.-OK
2.When reaching the endpoint, a command to reverse the spindle (e.g., from clockwise to counterclockwise).-OK
3.Continued synchronized motion beyond the specified end coordinate until the spindle actually stops and reverses.-OK
4.Continued synchronized motion back to the original coordinate.-OK
5.When reaching the original coordinate, a command to reverse the spindle a second time (e.g., from counterclockwise to clockwise).-OK, or offer user option for spindle stop or maintain /change direction? Lathe user may benefit from running reverse for gang tooling
6.Continued synchronized motion beyond the original coordinate until the spindle actually stops and reverses. Why?? Why not drop synchronization at start point?
7An unsynchronized move back to the original coordinate.-Bounce

I would see a point for 6 if peck tapping is in the works with full retract?

I see your point. It doesn't make sense to have a tapping motion that is synchronized in one direction and not in the other. It would be reasonable to end the synchronization when reaching the start point.
But what shall it do then? The movement has to end in a well defined position. If that is the start position this will mean fast brake, fast forward to start position so it will be a small bounce anyway. Another position could be specified I guess. I have no idea what the thoughts were when implementing the rigid tapping cycle, perhaps there were some good thoughts, .Idk.