If switch from high interval to low interval, the point will be inaccurate. Because the date of the point is not accurate. For example, the time will be round to 00:00 in 1d interval. In the same interval, the points should not be moved.
That makes sense, the only problem with this is...that you can't go back more than 900 periods at your chosen (lessor) interval. Therefore, making any line you draw (at a larger interval with a larger sampling) wildly inaccurate at a smaller time frame.
I understand it is a problem that you can't overcome with the current system, but it seems to be a real limitation.
I guess the only solution possible is the advent of readjusting the line at a smaller interval to match the actual data point in a smaller time frame. Since you are working on the next revision of bitcoinwisdom it seems that it would not be worthwhile to work on this in the older version.
I just hope the next system has some kind of compensation for this flaw. Because ultimately you can't be sure where the line actually should be at a higher frequency of samplings.
==================
Any idea if you could ever implement a Gann line system like the link above? Since you mentioned the line tool is actually lots of smaller lines I thought perhaps this might not be so difficult.