This appears to be fixed in 5.15, though the result is still not fully
simplified (which is a bug):
res: expand([t1,tt1]+t2) => [-3^-(1/2)/2, -1/(2*sqrt(3)) ]
expand(res) => [ -1/(2*sqrt(3)), -1/(2*sqrt(3)) ]
By the way, when pasting Maxima sessions to email, e.g. when reporting
bugs, please use display2d:false, because 2d output is not reliably
legible in email.
-s