I would like to record 48Khz/32bits.FP; but this message:
"...ALSA: final selected sample format for capture: 16bit little-endian..."
states that Im not getting 32bitFP signal incoming, isnt it?
How could I do to get 32bitFP signal at recording?.
The value reported as 16-bit is the resolution of your sound card's A/D converter. The incoming signal will be 16-bits, but Jack processes EVERYTHING at 32 bit floating point or greater, so this is converted by default (if you have a 64-bit machine, it will be 64-bit float).
Of course, it is on you to use a preamp to make sure you're getting the full dynamic range of your 16-bit card...you can decrease the quality significantly just by feeding your computer a weak input level.
That being said, you will be recording 32 bits if you have succeeded at getting JACK to start.
As you may know, the reason for 32 bit floating point is more important for any signal processing you may be doing as it improves the accuracy of filters, etc...so if you weren't using any plugins or doing any processing on the signal, this would be irrelevant, since you can't improve the accuracy beyond what you get from the sound card.
About Guitarix, I do understand what you mean about the treble not being as sharp. It may be all you need to do is use the convolver with a good amp cabinet response. I think the built-in cabinet simulator is EQ-based, but the convolver will give you as good of a cabinet response as any impulse you have recorded (for example, if you hijack one of the IR files from a commercial amp-sim, you will get equal results (for the cab response). The distortion/preamp tone is probably limited by the model they use...but I guess it's a matter of taste. Amp sims are getting better, but for the whole experience of a real amp, you just can't replicate it.
However, from the perspective of audience/listener, the difference will not be noticed, so it's really us picky musicians who care. A good guitar player sounds good even when using crappy gear that would disgust most people...anyway...