Initials (J) descending towards clef of hymns

[Matthew Roth]

is what I would like to have happen with initial letters.

I'm running latexmk (so it runs 2+ times as needed the first time, then at least once on subsequent compilations); the first time that I typeset, and then on a subsequent compilation, I got what I wanted. On intermediate passes (where it was typeset 2x on each press of "typeset") it did something which I have noticed with "regular" LuaLaTeX: the J descends into the clef.

What I need to know is how I can fix this for scores where it may be a problem.

 Hymns pose the most problem: Jesu Redemptor Omnium, now this Jesu Corona Virginum… probably as only the mode number appears in the annotation. Copying the classical Solesmes style (e.g. 1. Ant. as the first annotation, then 1. G as the second) fixes it for antiphons (however, as you can see, there's more room for this problem to occur with such an antiphon versus a hymn.

I should note too that I was typesetting to fix things that shouldn't have led to adjustments of the line of the hymn, just fixing the gabc of other chants where the line placement looked the same in the end.

% !TEX TS-program = lualatexmk
% !TEX parameter =  --shell-escape

%%these are TeXShop and mac-specific versions…

\documentclass[11pt]{article}

\usepackage{fontspec}
\usepackage[autocompile]{gregoriotex}

 \NewDocumentCommand{\gscore}{O{}mm}{%
 \greannotation{#1}%
\greannotation{#2}%
\grechangestyle{initial}{\fontsize{28}{28}\selectfont}%
 \gregorioscore{partitions/#3}%
 }

\setmainfont{EB Garamond}[
%UprightFont=EB Garamond Regular,
%ItalicFont= EB Garamond Italic,
%BoldFont= EB Garamond Bold,
%Ligatures=Rare,
Numbers=Proportional,
Numbers=OldStyle] %%has some odd effects on the long Q, but I use the local version

\begin{document}

\gscore[]{7. a}{an--joseph_fili_david--solesmes}

\gscore[]{4.}{hy_jesu_corona_virginum_in_tempore_paschali_solesmes_1961}

\end{document}

Having run that: