I have also done that. Running even just the preamp and no power amp sims directly into the powersection of the 5150 isn't even close at all. It actually sounds much closer just running through the matrix and utilizing the power amp section of the axefx and a/bing the two trough the cab. Which seems odd to me but actually seems to work better. The presence seems to be way off from the actual amp. I usually run the presence about 6 on the real amp and end up running it about 1.5 on the axefx to match up. I was just asking a question about looking at it since I have compared it to 3 different 5150's now. If you say its exactly like yours thats what I have to go with and just tweak from there and get it as close as I can. Thanks for the input.
If you turn off the power amp simulation you must set the presence control to noon. If you don't it won't be close at all. FWIW I own two 5150s, a 6505+ and a 5150 III and they all measure and sound identical to the models.
Furthermore... here's a plot of the output response of my 5150 Block Letter vs. the Axe-Fx model:
The amp is magenta, the Axe-Fx blue. Shows just how accurate MIMIC is.