Are you sure this is a "leakage" problem rather than a phase
shifter problem? How are you making that call?
I would expect "leakage" to be a low and somewhat beam-
forming-independent baseline (certainly if you kept total
panel power fixed, while moving the "steering") if all elements
are indentical and only beam forming routing differs.
Full power, but where you didn't send it, quite, is probably
phase shifters * routing. What you see far from beam center,
probably leakage.
A thing to look for, is unshielded stubs and not-perfectly-equal
routing of differential RF (if any). Time / phase skew can ruin
the expected cancellation of differential signals' common
mode, "in the moment", producing little "error wavees" on
every cycle - in effect these are also a phased array, with
uncontrolled phase and summing to an unknown sub-beam
magnitude and direction(s).