No that is the intercept term ( the value of Y when X=0 ). What they mean is that it is the prediction for Y when X=0 which is never thus it has no interpretation by itself and this is of little interest because X can be = 0.0
you can not pick another term as the value of B1 is conditional on B0 as they were simultaneously estimated.
prediction when x=0 is outside the range of experimentation and thus can be suspect. http://autobox.com/dave/OUTSIDE.png
EDITED AFTER QUESTION...
If B0 is set 10.0 for example then minimize (Y(i)-B1 * X(i) -10.0)**2 . This can be directly solved by subtracting 10. from each value of Y to get YY and then use a regression package to estimate the model YY=B1*X essentially requiring the regression equation to go through the origin . Good software like AUTOBOX , a piece of software that I helped to develop allows that feature/constraint while (nearly!) everybody else does not.