median comparison for data distribution

[bluesky2702]

I use a lot of census based housing data, specifically home value data for a given geographic area in a grouped distribution as shown below. The distribution shows how many homes fall into various price groups - for example, 63 homes valued at $100,000 to $149,999 in this geographic area. (Yes, I know about the negative number for the 0-49999 group.) There are a couple of issues with the quality of the census data:

1. The farther you get from the actual census, time-wise, the more questionable the data becomes. The data shown below consists of 2015 projections.
2. Since the census data is based on the input of census respondents, it is less accurate than actual home sales data since most homeowners are guessing at the value of their homes.

The workaround is to obtain median sale price from the local multiple listing service. In the example shown below, the median sales price from the local MLS is $778,400, 19.6% higher than the census-derived median of $650,742.

How can I use the MLS median price to redistribute the frequency counts to more closely approximating real-world data?

		Freq
		2015		Bracket	Endpoint
	Total	5570
	0	-73		49999	49999
	50000	33		49999	99999
	100000	63		49999	149999
	150000	98		49999	199999
	200000	163		49999	249999
	250000	222		49999	299999
	300000	562		99999	399999
	400000	783		99999	499999
	500000	1549		249999	749999
	750000	1017		249999	999999
	1000000	1153		1999999	1999999
		5570
	Median home value	650742

<tbody>
</tbody>