Create or code problem?

Creates can be finicky, yes. But the fact that it's consistaitly spinning low on only one turn is weird. If your create was bad it would be inconsistaintly inconsistaint, instead of consistaintly inconsistaint. Maybe using sleep(), which uses floats, is not running right on the cbc with all the lag from interpreted code and hardware constraints. Try using msleep() instead perhaps?
But first I'd check the tires of the create (how worn are they?), charge the battery, and perhaps try it on another create (if you have access to one) or another cbc. Also have you tested the code in different locations and orientations of the create? It could be that the board is slippery there, meaning you lose power on spinup or something.

I doubt that sleep vs msleep is the issue. C programs are compiled, not interpreted, and the overhead of casting the floating point value into an integer value is negligible unless you're calling it repeatedly.

Using timing to figure out how far you've traveled is generally not a good idea, because it varies widely. Use the distance functions in the Create library to determine exactly how far you've traveled in millimeters or degrees. For even better accuracy, use a Create OI script with the "wait distance" or "wait angle" command.

hanks for the previous reply

EDIT: I just realized the vertical nonmoving acceleration is around 64 = 1G, so it can measure +-2 G at maximum values? Also is the measured value in linear correspondance to actual G's? So if 64 = 1G then does 32 = 0.5G and 127 = 2G?