This is a known limitation, and how recipes and matricks-phase currently works (v1.6).
To get an even phase spread on different selections, you will need to either:
Combine the matricks phase with a matching matricks grouping,
e.g. [group=4, phase=0-270] or e.g. [width=6, phase=0-300]
use the "0 thru 360" / "0 thru -360" buttons in the attribute calculators, the "360" button in the PhaserEditor or "At Phase Percent 0 Thru 360" / "At Phase Percent 0 Thru -360" in commandline.
Only these interactions will (in v1.6) trigger the special internal function that automatically calculate and apply the actual phase-range needed to give an even spread, instead of applying the specified range literally (giving first and last equivalent phasevalues, 0 and 360)