Hi Neils,  Unfortunately I am not able to access your web site and
see the pictures.  Thus this posting, like my previous one, is based
on the assumption that by "late", you mean late 1920s or '30s.

In a rebuilt player action you have to look for a lot of possible
problems not normally found in an untouched unit.  For example,
a valve button may be screwed in too far, causing the problem you are
experiencing.  Poor trimming of a recovered pneumatic may result in
leakage at the glue joint.  Thicker valve leather may result in
insufficient valve travel.  And so on.

Aeolian wind-motors normally run at fairly high speed compared to many
other makes.  They appear to gobble up a lot of air, when actually
their pneumatics are of smaller capacity than the slower turning ones.
So you need the higher speed for adequate torque at the take-up spool.
The sprocket on the motor should be noticeably smaller than the one it
drives on the transmission -- just the opposite of the "Standard" brand
player action, for example.  Some rebuilders will reverse the two
sprockets; in theory, this would make the piano easier to pump by
reducing air flow through the motor.  But to re-calibrate the tempo,
you have to reduce tension on the governor spring, thus compounding the
problem greatly!

A faster wind-motor does more than drive the largest rolls with
more-than-adequate torque all the way to the end.  It provides a more
even, more constant speed, plus a faster de-crescendo after accenting
a note or chord.  (Late 20s/30s Aeolian foot-pumpers also have smaller
reservoirs for this last purpose.)

Regarding the more recent Aeolian plastic unit valve actions introduced
around 1960, the first of these, in spite of all their design flaws,
were remarkably easy to pump.  That's because their valves were made of
wood with leather glued to both sides.  Hard pumping began with the
change to sponge neoprene for both the valves and pouch discs.

Jeffrey R. Wood