I've been following the various power train mod threads and trying stuff along the way.
So far this is where I am at:
Band Power with lowest band set for 5W Max power
80 5 15
40 10 20
30 10 20
20 9 19
17 10 22
15 10 18
12 8 16
10 6 16
Measurements should be taken as relative more than accurate absolute values as this is just using the power meter in my SWR meter but somewhat validated by measuring peak voltages across a simple diode cap meter (QRPlabs 20W dummy load).
Q90 replaced with sot23 version of 2n2222
C81 replace with 1.1n - see comments below
Q912, 911 replaced with MPSH10
R86 20uH added in series
R87, 88 add 330pf caps in parallel
Q92, 93, 96, 97 replace with 2n2222 (plastic TO92)
R941, 942, 911, 96 - add 330pf in parallel
R261, 262 - remove
T11 replaced with BN43-202 2:4 (primary:secondary) of 0.7mm coated solid copper
100pf silver mica across the primary.
Some comments about variations tried long the way but noting that some of this may be path dependent (ie may vary if order of mods changes).
C81 at 470pf reduced the lower bands quite significantly (peaking was in place for everything pre PA at this stage) - many incremental values tried to get the lower bands up. May be capable of further tweaking.
I had no luck with the MPSH10 at Q92 and friends.
Mods were focussed on even-ness across the bands rather than max power at any particular point.
Transformer does not get warm.
Have some 2n3866 on the way.
Larger values of the silver mica produced a more pronounced lump on the power profile in the mid lower of these bands.
Power on 160m is very low (2W).
Happy to take suggestions on what to try next.
I do not have equipment to test IMD at various stages of other more complex measurements.
Some weird peaking at 17M.