I just reviewed the code and there is no problem. But the issue you describe exist in weekly timeframe, you are right. Look at my picture, the Y coordinate (price) of the 6th candlestick is correctly computed below the low of the next candlestick (the good one where it would be drawn), while the X coordinate (date) chosen by the platform is the previous barindex. Noobywan seems to have the same problem in his recent post in the PRT 10.3 issue thread.
I sent a bug report to PRT this morning.