I'm trying to create a modulation matrix for my patch, where I can route a bunch of control signals (LFO, ADSR, etc) to modulate various parameters using a matrix mixer topology (all mod sources can be sent to all destinations, and all destinations can receive a mix of all sources). I am naturally going for the [throw~] and [catch~] object to create one summing bus per destination (16 in total).
But I run into some errors there:

throw~ MO_pulse-width_mod: vector size mismatch
sigcatch summingBus1: unexpected vector size

I have figured out this has to do with DSP block size conflicts, as I have [block~] objects of varying sizes in several sub-patches. However I cannot get my head around DSP block size handling in Pd, and even less how to solve my issue. I have tried putting a number of size-matching [block~] objects on the subpatches containing the [throw~] and [catch~] objects and/or, where the source signal is actually being generated. None worked.

However, I noticed that if I replace [throw~]/[catch~] by {send~]/{receive~] and put a [block~ 64 1 1] in both sub-patches, the conflict is resolved. But using {send~]/{receive~] would greatly complexify the patch so I'd rather stick with [throw~]/[catch~] if I can.

I have tried to draw an accurate hierarchy of my current patch in the hope you may help me figure out where the mismatch occur. It's a tad complex as the modulation sources, the matrix mixer and the modulation destinations are all in different places of my patch. So I'm first using [s~] objects to send the various mod sources to the mod matrix, then use [throw~] objects to send these to the various "destinations" summing buses. I have added info where the signal generators are (~), [s~]/{r~] and {throw~/catch~] are all located in the hierarchy, as well as where some mandatory [block~] objects are (I cannot remove these as it will break my code).
Here it is:

Hope my post and this graph somehow makes sense...

It seems to work and I've no more error messages, which is good. But there's a very strange new glitch in that the catch~d signal is somehow twice the amplitude of the one being throw~d (it oscillate between -2 and +2). Why would that be in your opinion?

EDIT: got it. My reblocked parent patch needed to be [block~ 256 1 4]. I gather the signal was sent twice between each new blocking of the receiver, hence appearing as if 2 identical signals were being mixed.

I hope changing this to 4 won't somehow break the code I had running already. Will have to do some heavy testing now to make sure the rest of the patch still behaves correctly.