Constant in linear regression

Hey guys,
Need help in Linear regression:
I have a dataset of 70 tumors that during resection we used an electric measurement to estimate a distance to a certain point and than compared it to the post-operative imaging to the true distance
I have 2 continuous measures: 1.treshold (measured electrically) 2.distance (the true distance).

The correlation is 0.246 p=0.04.
Now, when I do linear regression - if I leave constant in (default) threshold is not significant at all with R^2 = 0.028
If I exclude constant, I get a very significant threshold p<0.0001 and R^2=0.58. which sounds too good to be true.
What is going on?

When threshold is 0 the distance should be 0, so leaving constant out seems logical, but then again, too good to be true.

Thanks in advance for explaining :)