It sounds like you might be measuring either the XTA outputs or the XTA + LF driver on its own, and not the whole speaker.
The loudspeaker’s overall response is a combination of both the processor, the loudspeaker driver characteristics and their relative physical position in the cabinet (w.r.t. delay/phase). So to flatten the phase of a loudspeaker, you need to measure the loudspeaker itself with all processing enabled.
(If you try to measure and correct the individual XTA outputs, you’ll be undoing the processing – particularly the crossovers – that is/are necessary to align the drivers. Also the loudspeaker dispersion, through the crossover frequency range, is controlled by the existing processing, and so if you change individual driver processing, you’re potentially changing the dispersion characteristics of the cabinet.)
When you have a measurement you’re happy with, take look at
and the “FIR Designer 3 | Overview” video at
(The video also discusses the filter delay. Maybe also take a look at https://eclipseaudio.com/fir-filter-guide/.)
Linearising a loudspeaker (in mag, phase or both) from a single measurement can be problematic. Whilst the phase might be what you want at that particular measurement point in space, the phase response in other directions could be very different. Generally this kind of thing should be done by taking measurements at many different angles (within the loudspeaker coverage pattern) and averaging the measurements (and preferably all in a clean measurement environment or anechoic chamber). FIR Designer includes a measurement average function – see the Overview video mentioned above.
Most of what we’re discussing here goes beyond the scope of how to use our software and relates more to the fundamentals of loudspeakers. These forums are mostly for how to use our software but we’ll help where we can.