Well that was a given in my question.
My thinking was that copying specific folders from the drive first like Documents and Pictures might result in less stress on the drive than reading all sectors from the beginning of the drive, presumably copying many gigabytes of useless OS and program files before getting anywhere near the user data.
It's been pointed out that mounting the drive and reading say 10GB of user files at the filesystem level might put more stress on the drive than reading say 1000GB sector by sector sequentially without mounting. It's a tough call and I'm not sure how anyone can be sure of the strategy as a blanket rule.
The other part of my question (not yet addressed by the other responders) was about the time and effort required to image a drive before attempting recovery of files. In my repair shop I'm not sure it's feasible to image every drive that I suspect is failing. I think the more practical strategy should take into account the majority of cases (in my experience) where files are easily recovered if a drive can be mounted, and drives dying completely in the middle of file recovery is rare. My usual charge is 1.5 hours labour for replacing a drive, re-installing Windows, all updates, installing a few basic apps, and attempting data recovery. In my market (home and small business users) it isn't feasible to spend that much time in every case.
My best practice is to image any drive that is questionable/has problems, no exceptions. The added advantage is a proper image process can include a log file so you can clearly see if there are bad blocks. Pretty simple concept. Even one bad block is reason to not bother doing anything else given the price of new or used HD's. Now that you have an image you can, after making a copy of it of course, work on getting data, doing P2V, etc.
But I don't do much in the break/fix arena these days. So I can understand someone wanting to shave time to improve turnaround times. Even over the native interface, such as SATA, this can take time for large drives. The risk is their's to make.
As mentioned you should use some Linux flavor to do this work. You can use the
diskpart command to prevent automount in Windoze. But I've never tried that. One reason is Windoze update has a habit of resetting system settings to default, automount enabled is the default setting. The reason I avoid using Windoze is, if I'm not paying attention, chkdisk may kick in if it detects a mountable partition with the dirty bit set.
What to charge the customer? Depends. Take a look at data recovery prices from various vendors and you will see they are usually flat rate, ignoring disk repairs. Personally I charge $50 for a setup fee, which is waived if I do a paid data recovery. I charge based on my time not elapsed time on a machine grinding away. So it's 1-2 hours. This is over and above any charges related to repairs.
A recent example. OS X Server, which uses mdadm for software raid. The internal drive was mirrored to an attached USB drive. Something poisoned the array so not only was it not booting to an OS with either drive, neither drive was working properly within Finder to recover data with in OS X. Imaged the drive using R-Studio and was able to get all the important stuff.
The data recovery part took almost 12 hours but I was using a USB2SATA bridge for the internal drive. Probably spent a total of 20 minutes monitoring the process. I charged 250 for the data recovery, drive and restore. Normally this would have been something like $350, $100 DR plus repair, but I cut them some slack as the total bill, due to onsite time, was well north of $1k.
