CFRM and Fictitious Play are very similar iterative algorithms to find Nash Equilibria. It is not true that CFRM only gives pure strategies, no. (The "current" strategy profile tends to have more pure strategies than with FP, but not the averaged final solution.)
Regarding speed:
Really depends on the game and the variant of CFRM. The Fictitious Play implementation in HRC is heavily optimized for holdem, so it tends to be significantly faster than unoptimized "Vanilla" CFRM for now. But CFRM seems to converge better in situations with many mixed strategies, e.g. with multiple bet/raise sizings.
Exploitability % while running:
Calculating exploitability is extremely fast for Fictitious Play and displaying it doesn't slow down the calculation. For CFRM it would take the runtime equivalent of a few iterations to update the exploitability, so i'm not sure if it's worth it.
Regarding the errors:
The errors indicate that some consistency checks for the results failed, which happens if the hand was modified while the UI results were calculated. These are normal if you were displaying/updating the hand while CFRM/FP was running, and once the calculation is stopped the errors should no longer occur. If no calculation was running, please try to re-produce the issue and upload the hand via "Hand: Share Hand".